GhostBus Hub is the marketplace for agent skills, built on GhostBus, the agent to agent message bus and shared workspace. 29 MCP tools, MIT, zero dependencies, on npm and the official MCP Registry. Humans browse here. Agents will pull from the registry underneath. Card checkout is open — every paid skill buys from its listing via Stripe, and Cash App still works.
Free core forever if you self host. Paid skills and hosted workspaces keep the lights on.
One studio, building in the open. The first product you meet on this site is the GhostBus Hub: a marketplace for agent skills where every listing is scanned, dated, and hashed before it goes live.
Ghost Developer Studio is a one-person studio in Oakdale, Tennessee, building developer tools, apps, and AI systems since 2022. Everything the studio ships carries its record with it: what it is, what state it is in, and how to check it. The product list below uses three plain labels, and the labels mean exactly what they say.
The GhostBus Hub is the studio's storefront, and the first product on this site. It lists 200 agent skills today: 1 free and 199 paid, at $12, $15, or $19 each, across 31 categories and 1,826 bundle files in total. Bought separately, the paid skills come to $2,701. Every listing shows its full file list, a preview taken from the skill itself, a worked example, a dated vetting record, and a bundle hash anyone can recompute. What that vetting record means is spelled out on the What vetted means page.
Paid skills are bought here on the storefront. Today that means Cash App: pay $RyanCotten2 the skill's exact listed price, put the skill name and your delivery email in the payment note, then email [email protected] from that same address so the payment can be matched. Each payment is confirmed by hand, and the skill is then delivered by email: the bundle plus a license key. The license is permanent, personal, and non-transferable, and it covers the version you bought. Refunds are goodwill, per the Hub terms published on this site. Card checkout is open now: the Buy button on any paid skill opens secure Stripe checkout at the exact listed price, and Cash App Pay is accepted there too.
Every skill here is first party for now, written and maintained by Ghost Developer Studio. Third party seller terms are published on this site as final v1.1: the Hub keeps a 20 percent commission and the seller receives 80 percent, and every submission is scanned and reviewed before it can list. Public submissions open at launch.
A member of the public can reach it now.
Built and proven on the studio's own machines, not public yet.
Not launched yet, whether it is still to be built or built and waiting on a launch decision.
The studio's marketplace for agent skills: 200 vetted skills with public file lists, previews, worked examples, and bundle hashes.
An agent message bus and MCP server, published on npm and listed on the official MCP Registry.
A governed browser tool for agents, with approvals, a provenance trail, and replay; published on PyPI and listed on the official MCP Registry.
An on-device, goal-driven agent app for Android; the code repo is public on GitHub.
A governed tool-call layer that sits between an agent's decision and the tool call; version 0.3.2 is published on PyPI.
A context slicer that shrinks big codebases before they reach the model; public repo with published releases on GitHub.
A load securement photo check: a photo goes in, and eight checks from the 49 CFR 393 rules are closed honestly as pass, check this, or could not tell. The working app is live on the web.
A hosted relay that lets agent tools on different machines share tasks, notes, and context; the relay is live on the web.
An offline code editor for Android and the web; release v3.10.1 is published on GitHub.
A family of small guardrail tools for agent work, published as 11 public repos on GitHub at v0.1.0.
Self-hosted email for the studio and its agents, with no ungated send path: mail is drafted, approved, and only then sent. Built and proven on a local test setup; not launched.
A phone service for agents on a zero-dependency SIP switch, with an approval gate on outbound calls. Built and proven locally; real-world numbers wait on carrier signups, so it is not launched.
A live-chat and shared inbox product, rebranded from Chatwoot. It runs today as a local sandbox instance; customer deployment is separate and has not launched.
A persistent, governed operator built on the Ghost Kernel, with a kanban dashboard. App v1.9.0 is delivered and in device testing; it is not on any store.
A provider-neutral kernel for governed agent systems, with a world model, an event graph, and permissioned actions. Build phases 1 to 10 are complete; public go-live waits on database provisioning.
An operating system experiment run by the Ghost Kernel. Alpha builds are delivered; a full trial day on a real phone is still ahead before wider work.
A client-side video editor. The Android build v1.0.1 is delivered and confirmed working; it sits in a private repo and has not been publicly released.
A language tool with a cloud beta in build. The beta build phase is done; the real beta has not run yet.
A governed hosting tool. The internal MVP is built; it is not deployed anywhere public.
A cyberdeck design and fabrication studio. The Android build is delivered; the public backend launch is still pending.
A developer dashboard. It is built, and its deployment is parked until the studio picks it back up.
The Hub's next step: open third-party seller submissions and run the full Hub on the studio's own server. Seller terms are already published as final v1.1 (the Hub keeps 20 percent, the seller receives 80 percent). The launch waits on a server decision, so submissions are not open yet.
An agent context and prompt optimizer for the studio's toolchain: a context tree, a token heatmap, and a one-click context export. Decided as a concept; not yet built.
Questions about any product on this page, or about selling on the Hub, go to [email protected].
Agent Seatbelt Guardrails is the Hub's free skill: a deterministic policy gate between an agent's decision and the tool call. It works with any host that supports the SKILL.md format.
What's inside
SKILL.md, examples/first-run/WALKTHROUGH.md
Preview, from the skill itself
“Agent Seatbelt is a deterministic policy gate between an agent's decision and the tool call. It classifies each Bash/PowerShell/file/MCP call, decides it against locked rules plus the project overlay, and writes an audit record BEFORE the call runs.”
Worked example, from the package
examples/first-run/WALKTHROUGH.md
Worked example: the first run. This is the first run the skill itself prescribes, done for real on Nov 11, 2026 against the real gate (ghost-seatbelt 0.3.2, seatbelt_hook.py): one dangerous call staged, its verdict read, and its record found in the audit log.
npx skills add littlestjames82-sys/ghostbus-hub-skills
Tested end to end Nov 11, 2026. The same skill and the full catalog also live on GitHub.
Agent Seatbelt Guardrails installs in one command and shows you how the Hub's gate works before you spend anything.
Answer by the problem you have, not by browsing two hundred listings.
Every bundle hash can be recomputed by anyone with the published procedure, no account needed.
Badges are earned from real actions on the Hub's own record. They are shown on the member's account, and they are never bought.
Accounts and badges live in the Hub app and open to the public at launch. Joining the launch list today is the first badge a visitor can line up.
Nov 10, 2026: six new skills, in wave 15.
Every agent in the fleet is judged on its own record: the cost for one period set beside the work that agent actually delivered, ending with one recorded call per agent, keep, narrow, or retire, with the reason in plain words.
The public launch day runs from one written runbook: the checks before the announcement closed on evidence, posts and pages landing in a set order, a named watcher on every surface, and rollback and correction lines approved before the first post goes out.
Archived files are tested on a set cadence against the archive index: each one opened in full and its hash compared with its index entry, so every failure is listed per file and can be repaired from the surviving copy while a good copy still exists.
Lapsed customers are reviewed one by one from their own job records, and a plain win-back note is drafted for each from what that record supports, held for your approval before anything is sent, with the answer recorded either way.
The coming year runs against a written plan built from your own prior year records: a planned figure and a reason for every income and expense class, totals worked in the open, and set review dates for reading the plan against actuals while the year can still be steered.
A workflow is planned end to end on one page: each step matched to a skill the operator holds or to a named gap, and the gaps ordered by what blocks the most work, so the cover held and the gaps that matter first are both plain to see.
the starter pack that keeps an agent inside rules you set.
two scans and a full listing record before you install anything.
polite, firm follow-up sequences that get invoices answered.
honest review asks sent at the right moment.
Small on purpose. Each skill below was checked with the Skill Vetting workflow before listing, and carries a vetted badge and a hash. Search and filter work right here in your browser.
Tip: press / to search
Showing 200 of 200 skills
Your agent stops asking permission for the safe calls, and can never quietly make the dangerous ones.
Vetted Nov 11, 2026 · full record in Catalog
Listing record: 8 of 8
A generic policy misfits your project: yours is written from your answers, every risky rule proven in check mode first.
Vetted Nov 4, 2026 · full record in Catalog
Listing record: 8 of 8
You know what a skill will do to your machine before it runs, in writing.
Vetted Nov 4, 2026 · full record in Catalog
Listing record: 8 of 8
Your repeated workflows become skills another agent can run without you in the room.
Vetted Nov 4, 2026 · full record in Catalog
Listing record: 8 of 8
Your agent drafts and prepares email all day; nothing leaves without your approval.
Vetted Nov 4, 2026 · full record in Catalog
Listing record: 8 of 8
Your agent works the call list you give it, briefed and approved first, and every call is logged.
Vetted Nov 4, 2026 · full record in Catalog
Listing record: 8 of 8
The next incident gets cheaper, because the last one was actually understood.
Vetted Nov 4, 2026 · full record in Catalog
Listing record: 8 of 8
Work passes between agents without the telephone game.
Vetted Nov 4, 2026 · full record in Catalog
Listing record: 8 of 8
Nothing ships on a green feeling; it ships on evidence, gate by gate.
Vetted Nov 4, 2026 · full record in Catalog
Listing record: 8 of 8
Your agent remembers what matters, and corrections actually stick.
Vetted Nov 4, 2026 · full record in Catalog
Listing record: 8 of 8
A backup you have already restored from once is the only kind that counts.
Vetted Nov 4, 2026 · full record in Catalog
Listing record: 8 of 8
Every inquiry gets a fast, honest reply, and you still send each one yourself.
Vetted Nov 4, 2026 · full record in Catalog
Listing record: 8 of 8
A quote from photos that says what it is built on, and what the photos cannot prove.
Vetted Nov 4, 2026 · full record in Catalog
Listing record: 8 of 8
Eight checks on a load photo, each closed honestly: pass, check this, or could not tell.
Vetted Nov 4, 2026 · full record in Catalog
Listing record: 8 of 8
One message each morning that earns its place, or a deliberate silence.
Vetted Nov 6, 2026 · full record in Catalog
Listing record: 8 of 8
Every figure you act on has been traced to the source that actually said it.
Vetted Nov 6, 2026 · full record in Catalog
Listing record: 8 of 8
A written scope a price can stand on, made before the work starts.
Vetted Nov 6, 2026 · full record in Catalog
Listing record: 8 of 8
Your agent watches the threads that matter and speaks only when something actually changed.
Vetted Nov 6, 2026 · full record in Catalog
Listing record: 8 of 8
The same agent, doing the same work, on a fraction of the context.
Vetted Nov 6, 2026 · full record in Catalog
Listing record: 8 of 8
Work you hand off comes back verified, not just reported.
Vetted Nov 6, 2026 · full record in Catalog
Listing record: 8 of 8
Finished work turns into paid work: the invoice is built from the job record, and the follow-up is already written.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Finished jobs turn into public proof, one honest ask at a time.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
You post at agent speed, and nothing goes out carrying a claim you cannot back.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Leads sit unanswered or half-known: every inquiry becomes a record, a readiness call, and one reply asking what is missing.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Quote numbers nobody can trace: each price sourced and named, and any unsourced price standing open instead of guessed.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Release notes users cannot use: every release written for its users, each line traceable to a real change.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
New agents start undefined: role and scope written down, tools granted one by one, gates set before the first task.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
A tangled backlog stalls everything: sorted into do now, schedule, delegate, and drop, with a written reason for every item.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Retiring an agent strands work and access: open work handed off, access withdrawn in order, records archived, nothing left dangling.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
An agent's mistakes have no limit: an owner approved budget logs each incident, and autonomy narrows as the balance falls.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Agents hold access beyond their role: inventory everything reachable, and every excess becomes a dated keep, narrow, or revoke decision.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Scattered secrets: each mapped by location and label into one proper home, every traveled secret rotated in a safe order.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Quiet quotes get a civil three-touch ladder to a clear yes or clean no, with every touch logged.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Job photos scatter unlogged: a before, during, and after record for every job, named and logged against the job record.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Materials runs waste trips: one planned run per job, quantities checked, items grouped by store section, substitutes agreed in advance.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Customers hear nothing until they chase you: five update messages drafted in your voice, from day-before reminder to finished note.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Callbacks argued from memory: the warranty promise sits in the job record, each callback checked against it, tracked to close.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Seasonal offers get improvised: one offer planned from your services, prices, and calendar, drafted to past customers first for approval.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Agent drift goes unnoticed: one dated health record of what is sound, what is not, and which fix comes first.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Charges appear with no cause: every charge traced to the agent and task behind it, against budget lines set beforehand.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Your agent meets real tricks untested: rehearsal drills against your own agent first, every finding turned into a standing rule.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Messy spreadsheets cannot be trusted: duplicates, broken rows, and unit fixes logged in a cleanup record, nothing dropped in silence.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Loose invoices, forms, and statements stay unsearchable: structured records filed under one naming rule, traceable to the exact page.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Folder you cannot find anything in: one planned pass, manifest approved before anything moves, undo log reverses the whole run.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Long recording sits uncut: a planned queue of captioned shorts, cut points traced to the transcript, posting order approved first.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Comments pile up unworked: each sorted answer, flag, or ignore, replies drafted in your voice, triage logged so nothing hangs.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Happy customers never get asked: the right customer asked once, at the right moment, in your voice, every ask logged.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Multi-job day planned on one sheet before anyone leaves: jobs ordered by drive time, crew and materials named per job.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Receipts vanish, profit rebuilt from memory: every receipt logged to its job's cost record, checked weekly against what was paid.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Your week's work stays unpublished: a newsletter drafted only from your own records, links checked, claims held to the records.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Meetings live in memory: notes become decisions and actions with an owner, a date, and the line each came from.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Every channel tells a different release story: each post drafted from one record, checked claim by claim, held for approval.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Drift arrives as emergencies: each scheduled pass leaves a dated report, findings triaged by risk, upgrades ordered one by one.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Data handling runs on habit: everything touched gets a class and a rule, public, internal, personal, or secret, violations flagged.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Tool results flow straight into work unchecked: shape and range checks plus a second source stand between result and action.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Claim rests on a link and memory: sources saved as full page text with address, date taken, and citation record.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Services live in your head: one menu draft, each service's includes and excludes plain, set prices shown, unset ones honest.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Done lives in memory: each finished job closes with a customer walkthrough, photos logged, warranty note handed over, sign-off filed.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Weather-killed day wrecks the week: affected jobs re-planned in promise order, crew knock-ons noted, customer messages drafted for your approval.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Customers go uncontacted: check-ins traced to each customer's job record, drafted in your voice, logged, never twice per period.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
One recording, no publishing package: show notes, chapter list, cleaned transcript, and description drawn only from what was actually said.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Jobs start on terms held in memory: deposit and materials rules set per job type, written into the quote wording.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Agents guess past their brief: work stops there, one escalation note carries the evidence and one exact question.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Big jobs return doubled, missing, or conflicting: split into work packages with hard boundaries first, merged against one shared record.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Consent lives in memory: one ledger per person and scope, checked before any details are used, withdrawals recorded and honoured.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Users report outages first: scheduled checks run user steps, every miss climbs a written retry, record, escalate ladder.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Guides describe tools you never shipped: every screen walked, every step written in the screens' words, checked by doing it.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Choices rest on figures nobody can trace: options compared on the same rows, every cell tied to named, dated source.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Closed jobs never show their profit: estimate versus actual cost, line by line, each lesson written into the next estimate.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Subs arrive with everything left to memory: scope note they agreed to, schedule and site rules in writing, payment recorded.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Wear surfaces mid-job: one maintenance record for every tool, vehicle, and machine, service intervals tracked, repairs logged at real costs.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Google profile drifts from the business: hours and services checked on schedule, photos from finished jobs, replies drafted for approval.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Posting days start blank: a stocked bank of topics from your real work records, each with its hook and record.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
New customers get different answers: one welcome pack built from your real records, services, terms, and what to expect inside.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Agent behavior changed and nobody can say when or why: every change versioned with its reason, earlier versions restorable whole.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Reports change shape every time: same shape each run, every figure traced to its record, missing records named as gaps.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Secret travel is invisible: every leg recorded with a risk call and a safer route, from labels and locations only.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Live incidents get improvised updates: each outward update drafted from the incident record in plain words, held for your approval.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Beta reports scatter: every report in one triaged record tied to its build, duplicates merged, severity called from its effect.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
CSV imports lose rows silently: load under an approved mapping, bad rows kept in a named rejects file, counts reconciled.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Day-of surprises on site: tomorrow's job checked today, access confirmed, materials counted, shortages raised while there is time to fix.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Distance priced from feel: your area becomes named zones, a written trip line per zone, stated identically in every quote.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Unhappy customers get argument: a ladder from first contact to resolution, each rung with its record, draft reply, and limit.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Jobs never become gallery: before and after pairs from each job's photo record, captions held to the record, permission noted.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Series contradict themselves: the arc and parts drawn from your work, continuity notes kept and checked before each new part.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Outreach built on numbers you cannot stand behind: a media sheet from your records, one pitch per prospect, all logged.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Workflows go live unproven: rehearsed against copies of your records, a step log flagging every outward or state-changing step first.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Task estimates are guesses: every estimate written before work starts, actual time beside it at the end, chronic misses named.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Handoffs break two steps later: every format written as a contract, fields and order fixed, every output checked against it.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Renewals land as surprise charges: subscriptions in one record with renewal dates and prices, last price beside each, keep-or-drop recorded.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Docs go stale: every claim in guides traced to the shipped build, drift logged by exact line, fixed or cut.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Forms carry figures from memory: every field mapped to a record first, values from that record only, filled copy checked.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Missed deliveries become arguments: day planned from job records, load and stop order written, customer message per stop, proof-of-drop noted.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Returns end in damage arguments: rented or borrowed items logged with owner, deposit, due date, condition noted out and back.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Week's money in is assumed, not proven: cash, Cash App, and card receipts matched line by line against job records.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Partner referrals depend on memory: one working list of who sends what work, what went each way, thank-yous kept current.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Live sessions start improvised: run order written from your real work, setup checked before you go live, follow-ups listed beforehand.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Filming lists ignore your audience: questions and requests from comments logged verbatim, grouped by theme, every item naming its comments.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Model or tool changes ship silent surprises: fixed task set runs on old and new setups, every difference named plainly.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Standing instructions go stale: every rule checked against the record of what happened, marked holding, stale, conflicting, or unproven.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Records kept forever by default: every type gets a written keep period and review date, deletion runs on schedule, logged.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Mistakes vanish in quiet edits: each fixed in place, then one plain-words log entry per correction is published and kept.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Retiring a product gets messy: decision and stop date recorded, sunset notice drafted for approval, records exported before anything stops.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Spreadsheet reports copy figures unchecked: every figure traced to the sheet's cells, totals recomputed, unsupported claims left out or flagged.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Morning starts with fetching: tools and materials staged tonight in job order, gaps flagged and filled before shops close.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Tomorrow starts from memory: every job closed out same day in one log, done, left, photos filed, next steps noted.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Credits argued from memory: every gift card, rain check, and store credit in one record, balances tracked redemption by redemption.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Reviews answered in hurry invent facts: each read fully, points checked against the job record, reply drafted, held for approval.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Title and thumbnail picks are last-minute guesses: variants recorded before publishing, then reviewed against the video's numbers and logged.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Guests run as scramble: prospects from your network tracked by stage, outreach drafted for approval, prep and follow-up logged.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
New agents reach real records unproven: what they can touch, send, and how they fail, proven on copies, call recorded.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Nobody can answer for the fleet: every agent's job, reach, and owner on one page, entries checked and kept current.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Procedure only one person knows: a dated runbook entry a second operator can follow at night, checked against the system.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Broken builds ship unchecked: one fixed pre-ship list for start, main flows, version stamp, and artifacts, record for every build.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Claims outrun their sources: every source graded before use, copies counted once, conflicts named in the open.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Library you cannot find things in: names, dates, and embedded fields agree under one rule, every change logged, originals untouched.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
First calls wander: a one-page script from your services sizes the job, only deliverable promises made, one next step agreed.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Supported add-ons missed: each one the job record supports written into the quote at list price, offered once, answer recorded.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Payment timing strains big jobs: billed in agreed stages, deposit, milestone midpoints, balance at completion, stages recorded open or paid.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Road costs vanish into recollection: every trip logged against its job, fuel stops recorded from receipts, true costs per job.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Crews start blind to hazards: one recorded walk names them across access, overhead, ground, and weather, each with a plan.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Payments blur together: every payment tied to its agreed stage, partials shown as partial, balance owed known from one record.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Questioned decisions settled from memory: every call recorded at the time with its inputs, rule, and outcome, repeating faults visible.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Fleet capacity gets guessed: one record per agent of load, queue growth, and the load where quality starts dropping.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Secrets age in silence: one rotation schedule holds each by label and location, with age, owner, next date, reminders ahead.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Pages ship with accessibility faults unseen: every page checked against a fixed list, findings logged, fixes drafted for your approval.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Releases ship license obligations: one ledger, rechecked against the build, states what each dependency carries and what its license asks.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Archived files cannot be found again: a searchable index built from each file's subject, date, parties, and type, kept current.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Rushes catch you unready: last year's job records become a season map with opening dates, honest slot counts, and waitlists.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Estimates quote last season's prices: most-bought materials rechecked on schedule from real listings, every check dated and sourced, changes logged.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Job labor costs come from memory: crew hours entered per job same day, totals per job and per week worked.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Vehicles roll unchecked: one standard check of lights, tires, hitch and coupler, load secured, logged per trip against the job.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Hand-typed records carry errors: every field gets a second pass against its source, mismatches listed by field, unreadable entries flagged.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
One name, conflicting entries: duplicates merged with every merge on record, stale entries flagged for your call, both values kept.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Incident gets two memories: timeline built from record and logs, what held, what failed, fixes owned and dated.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
One agent overloads, another idles: load read from the capacity record, moves proposed with reasons, approved before moving.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Pinned skills drift unseen: every package in a register, checked against its source on schedule, updates reviewed before anyone adopts.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Changes sneak into releases: one named candidate frozen, what is in and held out written down, exceptions approved in writing.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Screenshots mean nothing a month later: captures logged with what each shows, when taken, and which record it supports.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Conversions lose files in silence: batch runs under an approved plan, failures listed by name, counts reconciled, originals untouched.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Quotes lapse and return stale: each carries an expiry date, renewals rebuilt line by line from your current price record.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Cancelled jobs end in disputes: the deposit settled exactly as written terms say, every materials line accounted, refund worked openly.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Year-end gets rebuilt from scraps: income, expenses, and totals assembled into one packet, every total worked openly, every gap named.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Disposal costs land nowhere: runs planned from job records, fee from the site receipt, each job's share filed per job.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Warranty claims start from memory: proof of purchase, window checked openly, photos of the failure, claim drafted and tracked.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Repeat letters drift: each a checked template naming its purpose and required facts, filled from the record when sent.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Handovers start cold: brief carries state, work done, decisions, open items with holders and next steps, plus what to watch.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Approvals sit undecided: pending items on one record with age and owner, each ends decided, escalated, or expired by rule.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Reviewers get fragments, not the story: decision and proof trails exported into one packet, checked entry by entry, gaps named.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Release runs on memory: step by step against a written checklist, steps evidenced and signed off in order, holds named.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
People work from wrong copies: every document has one current version in a register, superseded copies marked, archived, changes logged.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Meetings start from memory: agenda built beforehand from the open items and decisions owed, each item with one owner.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Inbox is a pile: messages bucketed act now, file, waiting, or discard, counts add back to the start, nothing dropped.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Recurring jobs fail quietly: each on one schedule record with owner, cadence, and proof of last run, due misses marked.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Quotes rest on remembered prices: sell prices on one written list, reviewed on cadence, every change dated with its reason.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Endings stay open threads: the final account settled with arithmetic shown, records handed over dated, arrangements closed, closing note drafted.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Year-end becomes a document hunt: tax documents filed into one indexed folder as they arrive, monthly sweeps naming anything missing.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Praise gets lost or misused: reviews, messages, and comments gathered into one library, the permission state recorded on each item.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Multi-step jobs fail unnamed: whole job broken into ordered steps, each carrying its needs, owner, and a test for done.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Failures debut in front of clients: each failure scripted in advance, run on copies or in sandbox, observed behavior recorded.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Access drifts quietly: every grant in one standing register, rechecked on its due date, confirmed, narrowed, or revoked on record.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Minor slips share a queue with buyer-stopping faults: every incident graded against one written scale, grade and reasons on record.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Records stay past retention without explanation: each approved hold written as one exception naming records, reason, owner, and end date.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Workbook numbers rest on displayed values: every formula read range by range for errors, hidden typed values, and dead references.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Quotes carry stale terms: payment, changes, warranty, and validity blocks in one checked library, every new quote checked against it.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Who had which tool is guesswork: gear logged out to a named person and job, back in on one entry.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Invoices get argued from memory: disputed lines checked against the accepted quote and completion record, settlement recorded on the job.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Effort feeds sources that never send work: every lead recorded with its source, outcomes joined back on a review cadence.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Notes users cannot scan quickly: the change record rewritten in plain words, grouped new, improved, fixed, anything unsupported left out.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Questions lost: each kept in the asker's own words, where and when asked, grouped, counted, turned into FAQ drafts first.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Agent work gets judged by guesses: written account from the records, what went well or badly, lessons as dated notes.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Prompt changes become mysteries: each version kept whole in one registry with what changed, why, and which runs used it.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Access runs too wide: each tool scoped to least access per task type on one sheet, exceptions approved and dated.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Releases land on everyone at once: small first audience, one fixed list per stage, written widen, hold, or rollback call.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Tables trapped in PDFs: rows and totals lifted page by page, reconciled against printed totals, unreadable cells flagged, never guessed.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Names age badly: one written convention for files, versions, and deliverables, checker rules test each name before saving.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Paid slots get double booked: each booked against a named issue in a dated inventory, copy and link dates tracked.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Mod queue piles up: items read fully, decided on the rule each was flagged under, logged with the action taken.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Hiring runs on hurry and memory: role written from the season's real work, screening questions and pass marks fixed beforehand.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Cash lands unlogged: every payment logged the moment it lands, job, amount, and who took it, counted, reconciled against invoices.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Guarantees scatter: every one in a single ledger, what was promised, on which job, until when, claims tracked to resolution.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Job photos live only on a phone: batches counted, copied, reconciled, verified in the archive before any phone copy clears.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Handoffs break after brief changes: fixed scripted cases run them cold, one plain pass rule decides, failures logged by field.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
You learn records are wrong from customer: standing records sampled on cadence against written quality rules, error rates per rule.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Secrets shared in a hurry stay open: one named recipient, one purpose, single use or expiry, logged by label only.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Cover gaps surface during outages: every window carries a named holder and backup, handoffs and contact order written down.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Links rot in the docs: all links followed to their targets on cadence, dead or moved targets listed per page.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Terms render two ways: one glossary holds each term's approved rendering per language, new terms admitted only from a source.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Closed jobs teach nothing: compared estimate line against actual line on cadence, patterns of over and under named, adjustments drafted.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Sponsorships get judged from memory: each in one commitment log, promises and returns tracked both sides with dates and deliverables.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Rentals get double booked: bookings on one schedule with window, pickup and return times, and job, every change checked against.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Winterization varies by property: the same run each season, taps, pipes, heating, gutters in order, findings written on run record.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Quote photos go unmatched: chosen from the job's record, labeled to the quote line each supports, unbacked ones held back.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Captions change what videos say: every line checked against the spoken words before posting, errors listed per video and fixed.
Vetted Nov 9, 2026 · full record in Catalog
Listing record: 8 of 8
Every agent's cost goes unjudged: each judged on its record, cost beside work delivered, one keep, narrow, or retire call.
Vetted Nov 10, 2026 · full record in Catalog
Listing record: 8 of 8
Launch day runs on improvisation: a runbook with pre-announcement checks closed on evidence, posts in set order, rollback lines approved.
Vetted Nov 10, 2026 · full record in Catalog
Listing record: 8 of 8
Archive files rot untested: each opened in full on cadence, hash compared with the index entry, failures listed per file.
Vetted Nov 10, 2026 · full record in Catalog
Listing record: 8 of 8
Lapsed customers get forgotten: each reviewed from their job record, a plain win-back note drafted from it, held for approval.
Vetted Nov 10, 2026 · full record in Catalog
Listing record: 8 of 8
No plan for the coming year: figures and reasons for every income and expense class, built from prior year records.
Vetted Nov 10, 2026 · full record in Catalog
Listing record: 8 of 8
Gaps hide: each step matched to a skill you hold or a named gap, ordered by what blocks most work.
Vetted Nov 10, 2026 · full record in Catalog
Listing record: 8 of 8
Open directories list thousands of skills that anyone can upload. The record shows what that costs. In February 2026, Koi Security audited ClawHub and found 341 of 2,857 listed skills malicious (Koi Security, February 2026). Snyk scanned 3,984 skills across the open directories and found flaws in 36.8% of them, including 76 confirmed malicious payloads (Snyk, 2026).
The Hub takes the other approach. The catalog stays small, and it grows only when a package clears the same vetting every listed skill passed; it stands at 200 today. Every skill is first-party for now. Every package is scanned, with a human reading the findings in context before it lists. Every bundle hash sits on its listing, where anyone can check it.
Small is not the limitation here. Small is the security model.
| The Hub | Typical skill marketplaces | |
|---|---|---|
| Vetting evidence | Every listing publishes its scan record, file list, and bundle hash, dated. | A badge or a score, with the underlying record not published. |
| Checking a purchase | Paste the hash from your delivery email into the checker in the Help center and compare it yourself. | Trust the platform's word that the file is genuine. |
| License | Permanent, personal license. The version you buy is the version you keep, in writing in the buyer licence. | Terms that leave the version question open. |
| Refunds | Goodwill refunds decided by the operator personally, with the process written in the Help center. No blanket guarantee is promised. | A fixed window, decided by the platform. |
| Catalog | 200 skills, every one listed and priced on one page. $2,701 for the whole catalog. | Catalog size and totals not stated. |
| Worked examples | 200 of 200 listings carry a worked example you can read before buying. | Output samples on some listings. |
Every claim in the left column can be checked on this page.
As of Nov 10, 2026
Card and Cash App Pay are accepted at checkout.
Your skill bundle and license key are delivered by email. The listing's published bundle hash is the record of exactly what you bought.
Follow the install steps in the delivery email and on the skill's listing. The version you bought is the version you are licensed for.
Questions or a refund request: see the Terms section on this page.
The pull flow was built and tested end to end on a local GhostBus workspace on Nov 4, 2026, with no changes to GhostBus itself. The public hub is not live yet. Round 2 added license keys on top of the approval gate for both paid skills, tamper detection on pulled bundles, update checks against the registry, and a full smoke test of the hosted workspace lifecycle the paid plans will run on. Round 3 split the vault into its own private workspace on the hosted server with agent token enforcement on, closing the last procedural gate, and added the seller submission pipeline that scans every candidate package before a human reviews it. Here is what the local proof run did, using the bus tools agents already have.
The agent searches the Hub registry workspace on GhostBus and finds the skill, its version, price, and hash. No human needs to go hunting for a download link.
In the proof run, a buyer agent searched the registry, pulled the free Agent Seatbelt Guardrails bundle, and its sha256 matched the registry. For the paid test, a purchase task for Seatbelt Policy Authoring started in needs approval, with needsApproval: true. A claim before approval was refused. After human approval, the agent claimed it, pulled the 35 file bundle from the vault, and the hash matched. Buying becomes a governed bus action.
The pull is written to the GhostBus provenance log, and file history on the bus tracks new versions. You can see what was pulled, when, and which hash was verified.
The Hub runs its own mail service, GhostMail. It is self-hosted, with no rented inboxes and no per-seat fees. Each agent holds a mailbox tied to its bus name. Sending sits behind the same approval gate as paid pulls, and there is no ungated send path.
The Agent Commons is a live room where six agents with scripted roles perform real operations against real server state in real time: a planner, a builder, a reviewer, a buyer agent, a hub keeper, and an approver. The approver is a labeled stand-in for the human owner, not the owner.
The agents post messages, take on tasks, and run purchases through the approval gate. A purchase runs through the gate to a license and a hash-verified pull. The room trades in the three seed skills it was built with: Agent Seatbelt Guardrails, Seatbelt Policy Authoring, and Skill Vetting, not the 200 skill catalog sold on this site.
Post a message in the room and it becomes a task. When a purchase is waiting at the approval gate, a visitor can approve it and watch the license issue.
The room keeps one shared event log of everything that happens in it. The replay scrubs that same log, so what plays back is the room's own record.
GhostBus itself is free and MIT if you self host. Hosted is for people who want a managed workspace with a stable URL and backups handled. Pricing is per workspace.
For one project or one small team getting started.
For daily agent work where you need more room and faster help.
For studios and teams running several workspaces at once.
"Support is email, [email protected], and it is me personally, not a team. I aim to answer within one business day. Your workspace gets scheduled export backups, and you can export everything yourself anytime; your data is never locked in. There is no uptime SLA yet, this is a small, honestly run pilot."
Phase 2 opens the Hub to outside sellers at launch, and the system behind it is already built and proven locally. It stays manual and curated at first, so quality stays high with no payout automation to get wrong on day one.
The commission was decided on Nov 4, 2026: the Hub keeps 20% of the sale price and the seller receives 80%. On our own direct sales in this shop, we keep about $18.15 of $19 after payment fees. That is why this shop exists.
A third party sale ran end to end on a real hosted server, in the local test run:
The Hub sits on top of GhostBus. You do not need hosted to try it. Add this MCP config, point it at a shared file, and your agents share a workspace.
{
"mcpServers": {
"ghostbus": {
"command": "npx",
"args": ["-y", "ghostbus", "--store", "/path/to/shared/ghostbus-data.json"]
}
}
}
GhostBus is published as an MCP server you can discover through the registry, as well as directly from npm and GitHub.
Two hundred skills, each checked with the Skill Vetting workflow before listing. Every bundle on the Hub is hash verified against the registry: what an agent pulls is checked against the hash the registry holds, so a changed bundle does not pass as the listed one.
Checkout is open now: card via Stripe, or Cash App.
Not sure? Start with the free Guardrails skill and email [email protected] if you need a hand choosing.
These paths recommend an order. They are not bundles, and each skill is bought on its own at launch.
For a local service that lives on messages, photos, and site visits, this order takes a job from the first inquiry to a paid invoice and a public review.
The set: $73 bought separately, one skill at a time at launch.
For someone running agents every day, this order puts a gate in place first, then adds routine, watch, memory, and recovery.
The set: $51 bought separately, one skill at a time at launch, and the first step is free.
For someone who builds skills and ships work with agents, this order goes from writing, to checking, to releasing, to handing off, to posting.
The set: $83 bought separately, one skill at a time at launch.
Showing 200 of 200 skills
No skills match your search. Try a different word, pick another category, or reset the catalog to see all 200 skills.
Looking for something else? Email [email protected].
Your agent stops asking permission for the safe calls, and can never quietly make the dangerous ones.
What changes
Honest fit
For you if: your agent can call tools and you want the dangerous ones stopped cold, with the reason on record.
Skip it if: your agent only chats and never acts on anything.
Your first run produces: one real tool call decided by the gate, its verdict and reason in the log.
Preview, from the skill itself
“Agent Seatbelt is a deterministic policy gate between an agent's decision and the tool call. It classifies each Bash/PowerShell/file/MCP call, decides it against locked rules plus the project overlay, and writes an audit record BEFORE the call runs.”
Package makeup: 1 skill file · 1 file across 1 worked example
Vetting record
Scanned Nov 11, 2026 with the Skill Vetting scanner. 2 files scanned. Result: 2 leads, reviewed. One is the source link to the skill's own public GitHub repo. The other is the worked example's quoted audit log line, the hash-chained record the example exists to show.
Second scan (gitleaks 8.30.1, Nov 11, 2026): no secrets found.
Bundle sha256: a469a2603bce… Short form. The full hash lives in the registry and is verified on every gate run.
Bundle size: 10.0 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub The full package is published in that repository. Project repository: Agent Seatbelt on GitHub
Listing record: 8 of 8
What's inside
SKILL.md, examples/first-run/WALKTHROUGH.md
Worked example, from the package
examples/first-run/WALKTHROUGH.md
Worked example: the first run. This is the first run the skill itself prescribes, done for real on Nov 11, 2026 against the real gate (ghost-seatbelt 0.3.2, seatbelt_hook.py): one dangerous call staged, its verdict read, and its record found in the audit log.
Install the free skill with the standard skills CLI: npx skills add littlestjames82-sys/ghostbus-hub-skills (tested Nov 11, 2026).
A policy that fits your project, written from your answers, with every risky rule proven in check mode before it can block anything real.
What changes
Honest fit
For you if: you run the Seatbelt gate and your project has its own shape, its own tools, and its own lines.
Skip it if: you want a generic policy dropped in without answering the interview.
Your first run produces: a draft policy for one small project, and one rule proven in check mode.
Preview, from the skill itself
“Agent Seatbelt ships with locked rules that stop the universally dangerous things. What it cannot know out of the box is your project: which command deploys to production, where the real database lives, which destructive command is a normal Tuesday for you and which one would end the week.”
Package makeup: 1 skill file · 20 files across 5 worked examples · 9 references · 5 templates
Vetting record
Scanned Nov 4, 2026 with the Skill Vetting scanner. 35 files scanned. Result: 70 raw leads, reviewed at listing. They sit in the test batteries and pattern cookbook, the worked attack and policy samples this skill teaches operators to defend against.
Second scan (gitleaks 8.30.1, Nov 4, 2026): no secrets found.
Bundle sha256: da74e0f70ddf… Short form. The full hash lives in the registry and is verified on every gate run.
Bundle size: 107.3 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub The package itself is delivered with purchase. Project repository: Agent Seatbelt on GitHub
Listing record: 8 of 8
What's inside
SKILL.md, examples/ci-team/EXPLANATION.md, examples/ci-team/TEST_BATTERY.md, examples/ci-team/policy.json, examples/data-pipeline/EXPLANATION.md, examples/data-pipeline/INTERVIEW.md, examples/data-pipeline/TEST_BATTERY.md, examples/data-pipeline/policy.json, examples/harbor-api/EXPLANATION.md, examples/harbor-api/INTERVIEW.md, examples/harbor-api/TEST_BATTERY.md, examples/harbor-api/plan.json, examples/harbor-api/policy.json, examples/infra/EXPLANATION.md, examples/infra/INTERVIEW.md, examples/infra/TEST_BATTERY.md, examples/infra/policy.json, examples/policy-audit/AUDIT.md, examples/policy-audit/TEST_BATTERY.md, examples/policy-audit/flawed-policy.json, examples/policy-audit/policy.json, reference/ask-budget.md, reference/audit-taxonomy.md, reference/ci-and-team.md, reference/host-notes.md, reference/incident-to-rule.md, reference/interview-bank.md, reference/pattern-cookbook.md, reference/stack-archetypes.md, reference/variation-battery.md, templates/audit-scorecard.md, templates/plan.template.json, templates/policy-pr-template.md, templates/policy.template.json, templates/tuning-log.md
Worked example, from the package
examples/policy-audit/AUDIT.md
Worked example: the shopfront audit. shopfront is fictional: a small Next.js storefront, npm, deploys as npm run deploy:prod (and deploy:preview for previews)
You know what a skill will do to your machine before it runs, in writing.
What changes
Honest fit
For you if: you install skills other people wrote.
Skip it if: you only ever run code you wrote yourself.
Your first run produces: one real audit of a skill you were about to install anyway.
Preview, from the skill itself
“A skill is a set of instructions your agent will follow later, in your projects, with your permissions. Installing one on trust is the same as pasting an unsigned script into your own runbook because the description sounded friendly.”
Package makeup: 1 skill file · 7 files across 2 worked examples · 2 references · 1 script
Vetting record
Scanned Nov 4, 2026 with the Skill Vetting scanner. 11 files scanned. Result: 106 raw leads, reviewed at listing. The leads are the product itself, its pattern library and review checklist, plus a deliberately trapped example skill shipped as teaching material.
Second scan (gitleaks 8.30.1, Nov 4, 2026): no secrets found.
Bundle sha256: 8ca6bc269c6d… Short form. The full hash lives in the registry and is verified on every gate run.
Bundle size: 50.5 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/clean-skill/AUDIT.md, examples/clean-skill/SKILL.md, examples/clean-skill/reference/format-guide.md, examples/trapped-skill/AUDIT.md, examples/trapped-skill/SKILL.md, examples/trapped-skill/reference/quick-wins.md, examples/trapped-skill/reference/setup-guide.md, reference/report-template.md, reference/review-checklist.md, scripts/scan_skill.py
Worked example, from the package
examples/clean-skill/AUDIT.md
Skill audit: Meeting Notes Formatter (worked example). This is a fictional package, written as the clean half of this skill's pair of worked examples.
Your repeated workflows become skills another agent can run without you in the room.
What changes
Honest fit
For you if: you keep explaining the same workflow to your agent, week after week.
Skip it if: one-off prompts are all you ever need.
Your first run produces: a minimal skill skeleton that passes its own checklist.
Preview, from the skill itself
“A skill is a set of instructions someone else's agent will follow later, in their projects, with their permissions. That sentence is the whole discipline.”
Package makeup: 1 skill file · 3 files across 2 worked examples · 6 references · 3 templates
Vetting record
Scanned Nov 4, 2026 with the Skill Vetting scanner. 13 files scanned. Result: 0 findings.
Second scan (gitleaks 8.30.1, Nov 4, 2026): no secrets found.
Bundle sha256: 00fddc137435… Short form. The full hash lives in the registry and is verified on every gate run.
Bundle size: 30.9 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/from-idea-to-listing/WALKTHROUGH.md, examples/minimal-skill/AUTHORING_NOTES.md, examples/minimal-skill/SKILL.md, reference/description-triggers.md, reference/frontmatter-guide.md, reference/ground-rules-catalog.md, reference/listing-honesty.md, reference/self-vet-checklist.md, reference/structure-patterns.md, templates/authoring-scorecard.md, templates/example.template.md, templates/skill.template.md
Worked example, from the package
examples/from-idea-to-listing/WALKTHROUGH.md
Walkthrough: authoring Skill Vetting, end to end. A real package, authored with this workflow before this workflow was written down.
Your agent drafts and prepares email all day; nothing leaves without your approval.
What changes
Honest fit
For you if: your agent drafts or sends email in your name.
Skip it if: you want the agent to send without your approval. This skill will never do that.
Your first run produces: one real email, drafted and held at the gate, nothing sent.
Preview, from the skill itself
“Email is the agent channel most likely to cause a real world problem: it leaves the building, it carries the principal's name, and it cannot be unsent.”
Package makeup: 1 skill file · 2 worked examples · 5 references · 2 templates
Vetting record
Scanned Nov 4, 2026 with the Skill Vetting scanner. 10 files scanned. Result: 0 findings.
Second scan (gitleaks 8.30.1, Nov 4, 2026): no secrets found.
Bundle sha256: 81c3849041cf… Short form. The full hash lives in the registry and is verified on every gate run.
Bundle size: 19.8 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/inbox-triage/TRIAGE.md, examples/license-delivery/WALKTHROUGH.md, reference/addressing.md, reference/attachments-and-files.md, reference/gate-protocol.md, reference/inbound-handling.md, reference/license-delivery.md, templates/compose-brief.template.md, templates/triage-report.template.md
Worked example, from the package
examples/license-delivery/WALKTHROUGH.md
Walkthrough: a Hub license delivered by mail. The run this package's delivery pattern was taken from, end to end.
Your agent works the call list you give it, briefed and approved first, and every call is logged.
What changes
Honest fit
For you if: your agent makes or takes calls and you need a brief, an approval, and a log on every one.
Skip it if: you want high-volume calling without briefs or approvals.
Your first run produces: one briefed test call to a number you control, logged end to end, or the approved brief at the dial step.
Preview, from the skill itself
“Voice is the most committing channel an agent can use: it is live, it is personal, and the other side reasonably believes the voice speaks for the principal.”
Package makeup: 1 skill file · 2 worked examples · 5 references · 2 templates
Vetting record
Scanned Nov 4, 2026 with the Skill Vetting scanner. 10 files scanned. Result: 0 findings.
Second scan (gitleaks 8.30.1, Nov 4, 2026): no secrets found.
Bundle sha256: c7a1c0948144… Short form. The full hash lives in the registry and is verified on every gate run.
Bundle size: 19.3 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/inbound-call/INBOUND.md, examples/outbound-call/WALKTHROUGH.md, reference/call-gate.md, reference/extensions-and-contacts.md, reference/inbound-calls.md, reference/media-and-quality.md, reference/texting.md, templates/call-brief.template.md, templates/call-log.template.md
Worked example, from the package
examples/outbound-call/WALKTHROUGH.md
Walkthrough: an approved outbound call on GhostPhone. The run this package's gate pattern was taken from, end to end, on the studio's own switch.
The next incident gets cheaper, because the last one was actually understood.
What changes
Honest fit
For you if: an agent did something wrong, or nearly did, and you want it understood, not just fixed.
Skip it if: nothing has gone wrong yet and prevention is the job. Start with the free Guardrails skill instead.
Your first run produces: a one-page review of your most recent incident, with one finding tiered.
Preview, from the skill itself
“When an agent does something wrong, or nearly does, the first account of it is usually a story someone tells: reassuring, coherent, and built on memory. This skill produces the other kind of account: a review rebuilt from the records.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 4, 2026 with the Skill Vetting scanner. 9 files scanned. Result: 0 findings.
Second scan (gitleaks 8.30.1, Nov 4, 2026): no secrets found.
Bundle sha256: d6cb3fd3b032… Short form. The full hash lives in the registry and is verified on every gate run.
Bundle size: 19.3 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/injection-attempt-review/REVIEW.md, examples/stale-draft-near-miss/REVIEW.md, reference/corrective-actions.md, reference/evidence-and-inference.md, reference/severity-and-scope.md, reference/timeline-reconstruction.md, templates/incident-report.template.md, templates/timeline.template.md
Worked example, from the package
examples/injection-attempt-review/REVIEW.md
Worked review: the inbox instruction attempt (contained). A completed short review on invented but representative records, showing the tiers and the discipline at small scale.
Work passes between agents without the telephone game.
What changes
Honest fit
For you if: two or more agents, or two sessions, pass work between them.
Skip it if: one agent does everything inside a single session.
Your first run produces: one real capsule sent, and its receipt verified on return.
Preview, from the skill itself
“Most multi-agent failures are handoff failures: context that evaporates, scope that quietly grows in transit, two agents working the same job, a result nobody can trace back to its request.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 4, 2026 with the Skill Vetting scanner. 9 files scanned. Result: 0 findings.
Second scan (gitleaks 8.30.1, Nov 4, 2026): no secrets found.
Bundle sha256: e616a99f0187… Short form. The full hash lives in the registry and is verified on every gate run.
Bundle size: 18.9 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/bus-handoff/WALKTHROUGH.md, examples/refused-handoff/RECEIPT.md, reference/acceptance-and-verification.md, reference/capsule-anatomy.md, reference/failure-modes.md, reference/lifecycle-and-proof.md, templates/handoff-capsule.template.md, templates/handoff-receipt.template.md
Worked example, from the package
examples/bus-handoff/WALKTHROUGH.md
Walkthrough: a capsule handoff on the studio's bus. The pattern as the studio runs it, from a relay dress rehearsal in which a parent agent handed a subtask to a second agent through GhostBus and the studio's hosted relay, end to end.
Nothing ships on a green feeling; it ships on evidence, gate by gate.
What changes
Honest fit
For you if: you ship software and "done" keeps getting announced before it is true.
Skip it if: you never release anything anyone else depends on.
Your first run produces: one pending change run to its first gate verdict, with the release record started.
Preview, from the skill itself
“Releases fail in a characteristic way: not with a dramatic error, but with a quiet mismatch. The tag says one version, the package says another, the announcement describes a build nobody ran.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 4, 2026 with the Skill Vetting scanner. 9 files scanned. Result: 0 findings.
Second scan (gitleaks 8.30.1, Nov 4, 2026): no secrets found.
Bundle sha256: a45760b02556… Short form. The full hash lives in the registry and is verified on every gate run.
Bundle size: 17.9 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/held-release/RECORD.md, examples/version-truth/WALKTHROUGH.md, reference/evidence-standards.md, reference/rollback-and-recovery.md, reference/the-gates.md, reference/version-truth.md, templates/release-checklist.template.md, templates/release-record.template.md
Worked example, from the package
examples/version-truth/WALKTHROUGH.md
Walkthrough: a version truth reconciliation. From a workspace-wide improvement pass across the studio's repositories, condensed to the reconciliation itself.
Your agent remembers what matters, and corrections actually stick.
What changes
Honest fit
For you if: your agent works with you across weeks and keeps forgetting, or misremembering, what matters.
Skip it if: every task you give it is a one-off.
Your first run produces: three real memory notes, and one correction run the full distance.
Preview, from the skill itself
“An agent's memory is a promise to the next conversation: what is written there will be believed later, by a version of the agent that was not present when it happened. This skill keeps that promise honest.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 4, 2026 with the Skill Vetting scanner. 9 files scanned. Result: 0 findings.
Second scan (gitleaks 8.30.1, Nov 4, 2026): no secrets found.
Bundle sha256: b92354a292d5… Short form. The full hash lives in the registry and is verified on every gate run.
Bundle size: 19.2 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/correction-propagation/WALKTHROUGH.md, examples/parked-plan/WALKTHROUGH.md, reference/corrections-and-conflicts.md, reference/hygiene-and-forgetting.md, reference/provenance-and-trust.md, reference/what-earns-a-note.md, templates/correction-record.template.md, templates/memory-note.template.md
Worked example, from the package
examples/correction-propagation/WALKTHROUGH.md
Walkthrough: the phone number correction. A real correction from the studio's own memory, worked end to end.
A backup you have already restored from once is the only kind that counts.
What changes
Honest fit
For you if: you have backups you have never once restored from.
Skip it if: you already run restore drills and can show the last one.
Your first run produces: one real backup with its manifest, and one file restored and verified.
Preview, from the skill itself
“A backup that has never been restored is not a backup; it is a belief.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 4, 2026 with the Skill Vetting scanner. 9 files scanned. Result: 0 findings.
Second scan (gitleaks 8.30.1, Nov 4, 2026): no secrets found.
Bundle sha256: 2e38fe6cfce1… Short form. The full hash lives in the registry and is verified on every gate run.
Bundle size: 18.4 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/hub-restore-drill/WALKTHROUGH.md, examples/the-config-that-wasnt/LESSON.md, reference/drill-cadence.md, reference/failure-modes.md, reference/manifest-discipline.md, reference/restore-verification.md, templates/backup-record.template.md, templates/drill-report.template.md
Worked example, from the package
examples/hub-restore-drill/WALKTHROUGH.md
Walkthrough: the combined hub and mail drill. The drill pattern as the studio runs it against its own marketplace state and mail service, condensed from the tools' actual behaviour.
Every inquiry gets a fast, honest reply, and you still send each one yourself.
What changes
Honest fit
For you if: you sell on Marketplace and answer the same questions every day.
Skip it if: you want replies sent automatically without reading them. The operator sends here.
Your first run produces: one live thread drafted and held, ready for you to paste and send.
Preview, from the skill itself
“Local service work is won and lost in the message thread. A buyer asks three sellers the same question; the one who answers clearly, asks the right qualifying questions, and quotes honestly gets the job.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 4, 2026 with the Skill Vetting scanner. 9 files scanned. Result: 0 findings.
Second scan (gitleaks 8.30.1, Nov 4, 2026): no secrets found.
Bundle sha256: 7fced4111374… Short form. The full hash lives in the registry and is verified on every gate run.
Bundle size: 18.4 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/firewood-half-load/WALKTHROUGH.md, examples/valley-roof-repair/WALKTHROUGH.md, reference/drafting-in-voice.md, reference/money-and-terms.md, reference/thread-tracking.md, reference/triage-and-facts.md, templates/inquiry-record.template.md, templates/reply-draft.template.md
Worked example, from the package
examples/firewood-half-load/WALKTHROUGH.md
Worked example: the half load of firewood. A real-shaped inquiry from the studio's firewood listing: a buyer two towns over asks what a half load delivered would cost.
A quote from photos that says what it is built on, and what the photos cannot prove.
What changes
Honest fit
For you if: you quote local jobs from the photos customers send you.
Skip it if: you measure everything in person before you ever name a price.
Your first run produces: one honest worksheet quote from a real set of job photos.
Preview, from the skill itself
“Buyers send photos and ask 'what would this run'. The tempting answer is a confident number; the correct answer is a disciplined one.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 4, 2026 with the Skill Vetting scanner. 9 files scanned. Result: 0 findings.
Second scan (gitleaks 8.30.1, Nov 4, 2026): no secrets found.
Bundle sha256: e5c8a7af12e0… Short form. The full hash lives in the registry and is verified on every gate run.
Bundle size: 18.2 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/cleanup-lot-from-photos/WALKTHROUGH.md, examples/valley-from-photos/WALKTHROUGH.md, reference/ask-backs.md, reference/photo-evidence.md, reference/pricing-discipline.md, reference/quantities-from-photos.md, templates/photo-request.template.md, templates/quote-worksheet.template.md
Worked example, from the package
examples/cleanup-lot-from-photos/WALKTHROUGH.md
Worked example: the overgrown lot. A land cleanup inquiry where the photos are good and the quote still should not be firm, for a different reason than the valley: here the quantities behave and the unknown is a decision, not a measurement.
Eight checks on a load photo, each closed honestly: pass, check this, or could not tell.
What changes
Honest fit
For you if: you haul loads and want a second set of eyes on the photo before you roll.
Skip it if: you need a certified inspection. A photo check is not one, and the skill says so itself.
Your first run produces: one completed check report on a load photo you already have.
Preview, from the skill itself
“A loaded trailer photographed from four sides holds real evidence about how it will travel, and real limits about what it can prove.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 4, 2026 with the Skill Vetting scanner. 9 files scanned. Result: 0 findings.
Second scan (gitleaks 8.30.1, Nov 4, 2026): no secrets found.
Bundle sha256: ee561ea0011d… Short form. The full hash lives in the registry and is verified on every gate run.
Bundle size: 19.2 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/rear-overhang/WALKTHROUGH.md, examples/tarp-gap-and-hidden-strap/WALKTHROUGH.md, reference/limits-and-law.md, reference/photo-capture.md, reference/the-eight-checks.md, reference/verdict-discipline.md, templates/capture-checklist.template.md, templates/check-report.template.md
Worked example, from the package
examples/rear-overhang/WALKTHROUGH.md
Worked example: long at the back, nothing marking it. A flatbed of bundled lumber, photographed at a yard in daylight, worked as the method prescribes.
One message each morning that earns its place, or a deliberate silence.
What changes
Honest fit
For you if: your morning starts with five apps and a pile of notifications.
Skip it if: you have no recurring sources worth briefing from.
Your first run produces: yesterday's briefing, run today, in tomorrow's format, or a proven silence.
Preview, from the skill itself
“# Agent Daily Briefing
The briefing is the one message the agent sends that nobody asked for, so it has to earn its place every morning.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 6, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: 01a286c5a37b… Short form. The full hash lives in the registry and is verified on every gate run.
Result: no findings.
Second scan (gitleaks 8.30.1, Nov 6, 2026): no secrets found.
Bundle size: 25.7 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/a-morning-with-signal/WALKTHROUGH.md, examples/the-quiet-day/WALKTHROUGH.md, reference/format-and-timing.md, reference/ranking-the-day.md, reference/silence-rules.md, reference/sources-and-dedupe.md, templates/briefing.template.md, templates/source-checklist.template.md
Worked example, from the package
examples/a-morning-with-signal/WALKTHROUGH.md
A real-shaped morning for a solo operator, worked the way this package prescribes: eleven candidates in, two items out, and every strike accounted for. The renewal's two copies merge into one candidate naming both sources.
Every figure you act on has been traced to the source that actually said it.
What changes
Honest fit
For you if: you act on numbers you find, with money, safety, or a real decision riding on them.
Skip it if: the claim changes nothing for you either way.
Your first run produces: one claim from this week audited to a verdict, source named.
Preview, from the skill itself
“# Verified Research
A research summary is only as good as the weakest figure in it that somebody acts on.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 6, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: eae2f35be458… Short form. The full hash lives in the registry and is verified on every gate run.
Result: no findings.
Second scan (gitleaks 8.30.1, Nov 6, 2026): no secrets found.
Bundle size: 26.4 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/the-conflicting-claim/REVIEW.md, examples/the-land-record/WALKTHROUGH.md, reference/claim-audit.md, reference/people-and-property.md, reference/primary-sources.md, reference/reporting-the-dive.md, templates/findings-report.template.md, templates/research-brief.template.md
Worked example, from the package
examples/the-land-record/WALKTHROUGH.md
A listing offered "the Hartley parcel, Ridge Road, about 120 acres of woodland" in Alder County. "About 120 acres" per the listing. Verdict: contradicted. The parcel record says 103.4 acres, and that is the figure the report carries.
A written scope a price can stand on, made before the work starts.
What changes
Honest fit
For you if: you price local work on site and a verbal scope has burned you before.
Skip it if: you only take work that arrives with the scope already written.
Your first run produces: one real job rebuilt into a written scope.
Preview, from the skill itself
“# Job Site Scoping
A price quoted from photos is a promise made at a distance.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 6, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: 1f7d43bd52b2… Short form. The full hash lives in the registry and is verified on every gate run.
Result: no findings.
Second scan (gitleaks 8.30.1, Nov 6, 2026): no secrets found.
Bundle size: 28.2 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/shed-concrete-prep/WALKTHROUGH.md, examples/the-leaking-valley/WALKTHROUGH.md, reference/day-rate-or-job-price.md, reference/photos-and-measures.md, reference/the-walkthrough.md, reference/the-written-scope.md, templates/scope-visit-checklist.template.md, templates/scope-writeup.template.md
Worked example, from the package
examples/shed-concrete-prep/WALKTHROUGH.md
The shed measures 24 by 32 feet off the tape. Floor area is closed arithmetic: 768 square feet, and the gravel for a 4-inch base works out near 9 and a half cubic yards, stated in the scope as an estimate because the depth varies with the floor's humps.
Your agent watches the threads that matter and speaks only when something actually changed.
What changes
Honest fit
For you if: you wait on replies, postings, or changes that you currently check by hand.
Skip it if: nothing you care about ever changes outside your own tools.
Your first run produces: one live watch set on a real thread, its silence rules written down.
Preview, from the skill itself
“# Agent Watchkeeping
A watch earns its keep by the messages it does not send.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 6, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: 3dec99b45c4d… Short form. The full hash lives in the registry and is verified on every gate run.
Result: no findings.
Second scan (gitleaks 8.30.1, Nov 6, 2026): no secrets found.
Bundle size: 30.7 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/the-buyer-watch/WALKTHROUGH.md, examples/the-reply-watch/WALKTHROUGH.md, reference/baseline-discipline.md, reference/change-detection.md, reference/ending-a-watch.md, reference/reporting-changes.md, templates/change-report.template.md, templates/watch-brief.template.md
Worked example, from the package
examples/the-reply-watch/WALKTHROUGH.md
A solo operator pitches a collaboration to a video creator by direct message, then opens a watch on the thread: three checks of silence, one reply, and a watch that ends itself. The end conditions are two: a reply lands, or Friday 17:00 passes.
The same agent, doing the same work, on a fraction of the context.
What changes
Honest fit
For you if: your agent bills feel high and every task carries the whole project along.
Skip it if: your tasks are short and your context is already lean.
Your first run produces: one real task's context audited by section, with its slice plan.
Preview, from the skill itself
“# Context Cost Control
Every token an agent carries into a task is paid for twice: once in cost, and once in the attention it crowds out.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 6, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: 01c7673f0e43… Short form. The full hash lives in the registry and is verified on every gate run.
Result: no findings.
Second scan (gitleaks 8.30.1, Nov 6, 2026): no secrets found.
Bundle size: 19.9 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/the-false-saving/REVIEW.md, examples/the-oversized-brief/WALKTHROUGH.md, reference/measuring-the-context.md, reference/slicing-strategies.md, reference/summaries-with-provenance.md, reference/verifying-the-saving.md, templates/context-audit.template.md, templates/slice-plan.template.md
Worked example, from the package
examples/the-oversized-brief/WALKTHROUGH.md
A delegated task arrived with everything its author had ever touched attached to it. The figures below are this example's own, kept because the arithmetic is the lesson. Each section was marked: read by the task, decision depends on it.
Work you hand off comes back verified, not just reported.
What changes
Honest fit
For you if: you hand work to other agents and read their reports afterward.
Skip it if: you do every task yourself, in one session, start to finish.
Your first run produces: one small task delegated and verified by hand, the full loop run once.
Preview, from the skill itself
“# Delegated Work Control
Delegated work fails in a characteristic way: not with an open error, but with a confident report.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 6, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: a334f734ac13… Short form. The full hash lives in the registry and is verified on every gate run.
Result: no findings.
Second scan (gitleaks 8.30.1, Nov 6, 2026): no secrets found.
Bundle size: 20.2 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/the-clean-handoff/WALKTHROUGH.md, examples/the-inflated-count/WALKTHROUGH.md, reference/parent-verification.md, reference/reading-a-report.md, reference/sanity-gates.md, reference/writing-the-brief.md, templates/delegation-brief.template.md, templates/verification-checklist.template.md
Worked example, from the package
examples/the-clean-handoff/WALKTHROUGH.md
From a documentation job in the studio's own workspace, condensed to the checking itself. The parent works the six done items against the returned files, not against the summary.
Finished work turns into paid work: the invoice is built from the job record, and the follow-up is already written.
What changes
Honest fit
For you if: you do paid work in person, quote it up front, and collect after, and the collecting is the part that slips.
Skip it if: you bill through a platform that invoices and chases for you, or every job is paid in full before the work starts.
Your first run produces: one checked invoice, or the next follow-up owed, for your most recent finished job, every figure traced to the record, held for your send.
Preview, from the skill itself
“# Invoice and Follow-Up
The job is not finished when the work is; it is finished when the invoice is paid.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: 4ca84478d9e5… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 21.6 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/balance-after-deposit/WALKTHROUGH.md, examples/the-slow-payer/WALKTHROUGH.md, reference/building-the-invoice.md, reference/figures-and-changes.md, reference/the-follow-up-ladder.md, reference/the-job-ledger.md, templates/invoice-draft.template.md, templates/job-ledger.template.md
Worked example, from the package
examples/balance-after-deposit/WALKTHROUGH.md
A roof valley repair, finished on a Tuesday. The invoice is the interesting part, because money had already moved before the work started, and more work appeared once the roof was open.
Finished jobs turn into public proof, one honest ask at a time.
What changes
Honest fit
For you if: your work is good, your customers say so, and your public pages are quieter than your job book.
Skip it if: your work arrives through channels where public reviews never reach your next customer.
Your first run produces: a written call on your last three finished jobs (earned, fix first, or already asked) and one held ask, ready to send.
Preview, from the skill itself
“# Review Requests
A happy customer forgets to say so in public.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: 377e4a9ce1ee… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 21.4 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/the-fix-first-job/WALKTHROUGH.md, examples/yard-cleanup-review/WALKTHROUGH.md, reference/drafting-the-ask.md, reference/platforms-and-links.md, reference/the-ask-log.md, reference/the-right-moment.md, templates/ask-and-thanks.template.md, templates/ask-record.template.md
Worked example, from the package
examples/the-fix-first-job/WALKTHROUGH.md
A land-clearing and brush job for a couple, the Hartleys here: a back lot cleared for a garden, stumps ground, brush hauled. Midway through the second day, the operator's trailer clipped their gate post backing out.
You post at agent speed, and nothing goes out carrying a claim you cannot back.
What changes
Honest fit
For you if: you ship real things and post about them, or mean to, and want the posts fast without the exaggeration risk.
Skip it if: you post rarely and write every word yourself anyway; the gate earns its keep at volume.
Your first run produces: one checked draft post for the most recent thing you shipped, with its proof list attached, held for your approval.
Preview, from the skill itself
“# Social Posting Gate
Nothing posts that a human has not read.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: e183ec88115b… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 22.0 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/the-almost-claim/WALKTHROUGH.md, examples/the-release-post/WALKTHROUGH.md, reference/claim-checking.md, reference/platform-facts.md, reference/the-brief.md, reference/voice-rules.md, templates/post-brief.template.md, templates/posting-record.template.md
Worked example, from the package
examples/the-almost-claim/WALKTHROUGH.md
The gate earns its keep on the drafts that feel most routine. This is the other classic case: not a false claim invented, but a true plan announced a release too early.
No lead sits unanswered or half-known: every inquiry becomes a record, a readiness call, and the one reply that asks for exactly what is missing.
What changes
Honest fit
For you if: leads reach you across Marketplace, text, and email, and some sit half answered because nobody wrote down what was missing.
Skip it if: you want the agent to message leads or quote them a price. This skill builds the record and drafts one reply; the operator sends.
Your first run produces: one real lead as a complete intake record, with its missing list, a readiness call with reasons, and one held reply draft the operator can send.
Preview, from the skill itself
“# Lead Intake
A lead is not a job yet.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: 1cff687cf5fe… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 28.9 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/the-complete-referral/WALKTHROUGH.md, examples/the-vague-text/WALKTHROUGH.md, reference/readiness-and-routing.md, reference/the-clarifying-reply.md, reference/the-intake-record.md, reference/the-missing-list.md, templates/clarifying-reply.template.md, templates/intake-record.template.md
Worked example, from the package
examples/the-complete-referral/WALKTHROUGH.md
A message through the studio's email, Monday morning, from a past customer passing on a neighbour. Referrals are the best leads a local operator gets, and this one shows what a complete one looks like.
The customer sees where every number came from: each price sourced and named, materials, incidentals, and profit in their own columns, and any price that could not be sourced standing open instead of guessed.
What changes
Honest fit
For you if: you price jobs before the work starts and want every number on the estimate tied to a named source the customer can check.
Skip it if: you are comfortable quoting from memory in one lump total. Here the columns never blend, and an unsourced price stands open as a placeholder.
Your first run produces: one complete estimate held for the operator, every line sourced or openly open, every quantity traced or labeled, the totals re-added.
Preview, from the skill itself
“# Job Estimate Builder
An estimate is a promise written in numbers, and the customer reads it as a test of the operator: can this person show me where every figure came from?”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: 9084189437d7… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 29.5 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/the-price-that-could-not-be-sourced/WALKTHROUGH.md, examples/the-roof-valley-repair/WALKTHROUGH.md, reference/placeholders-and-honesty.md, reference/quantities-and-units.md, reference/sourcing-prices.md, reference/the-three-columns.md, templates/estimate.template.md, templates/materials-list.template.md
Worked example, from the package
examples/the-price-that-could-not-be-sourced/WALKTHROUGH.md
A porch floor repair with a debris haul, and one material that defeats the sourcing pass. The figures are this example's own, each from the source named beside it, exactly as the skill requires; they show the method, not a price list.
Every release reads like it was written for the people using the software, and every line in it can be traced back to a real change.
What changes
Honest fit
For you if: you ship versioned software and the changelog is missing, written in git words, or thrown together at release time.
Skip it if: you do not keep releases, or you want notes published without a human approving them. Every entry here traces to a real change.
Your first run produces: a draft changelog section held for approval, with its trace list beside it: every entry pointing at a commit or a note.
Preview, from the skill itself
“# Changelog Keeper
A changelog is the one document that tells the people using a piece of software what a new version does to them.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: 6e87d4d56052… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 24.0 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/the-messy-range/WALKTHROUGH.md, examples/the-patch-release/WALKTHROUGH.md, reference/reading-the-range.md, reference/sections-and-order.md, reference/user-words-not-git-words.md, reference/what-does-not-go-in.md, templates/changelog-entry.template.md, templates/release-notes.template.md
Worked example, from the package
examples/the-messy-range/WALKTHROUGH.md
Fieldbook again, this time the range for a bigger release: features, fixes, one breaking change, a pile of internal work, and one feature that did not survive the range. The commits below are invented for this walkthrough; the keeping is real.
Sets up a brand-new agent with its role and scope written down, tools granted one by one with reasons, limits and approval gates set before the first task, and a starter task list that proves the setup, all held in an onboarding record.
What changes
Honest fit
For you if: a new agent is starting real work, or an existing one is taking on a new role, and its setup so far is a vague role and a pile of tools.
Skip it if: you want the tools granted first and the limits worked out later. Here the role, the limits, and the gates are written before the first task.
Your first run produces: an onboarding record holding the role brief, the tool grants, the limits, the gates, and the checked result of each starter task.
Preview, from the skill itself
“# Agent Onboarding
Most agent trouble does not start with a bad task.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: 621872e31358… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 39.6 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/research-agent-for-a-small-studio/WALKTHROUGH.md, examples/support-agent-with-tight-limits/WALKTHROUGH.md, reference/limits-before-tasks.md, reference/the-role-brief.md, reference/the-starter-week.md, reference/tools-one-by-one.md, templates/onboarding-record.template.md, templates/role-brief.template.md
Worked example, from the package
examples/research-agent-for-a-small-studio/WALKTHROUGH.md
A small studio with four people takes on more client questions than it can answer from memory: what similar studios offer, what a tool costs, what a supplier's current terms are. The owner wants an agent to do the looking up and hand back short, sourced notes, so the studio answers from facts instead of impressions. This is how that agent was onboarded, before it answered a single client question for real.
Sorts a tangled backlog into do now, schedule, delegate, and drop, with a written reason for every item and no drop left silent.
What changes
Honest fit
For you if: your backlog has grown past what anyone can hold in their head, and old items vanish by being forgotten instead of decided.
Skip it if: your queue is small and current, or you want items dropped quietly. Every drop here carries a reason, a date, and an owner.
Your first run produces: a complete triage note held for the owner: every item placed, every drop carrying a reason and an owner, and the do now bucket small enough to finish.
Preview, from the skill itself
“# Task Queue Triage
A queue that only grows stops being a plan and becomes a guilt list.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: d09f323584a2… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 36.8 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/agent-backlog-busy-week/WALKTHROUGH.md, examples/shared-queue-after-launch/WALKTHROUGH.md, reference/dependencies-first.md, reference/staleness-and-drops.md, reference/the-four-buckets.md, reference/value-and-risk.md, templates/queue-review.template.md, templates/triage-note.template.md
Worked example, from the package
examples/agent-backlog-busy-week/WALKTHROUGH.md
An agent that runs errands, drafts, and checks for a small studio has let its backlog grow for six weeks. A busy week is coming, with client work on four of five days. The owner gives the pass one instruction: Monday morning should start with a short list that matters, not with forty items and a feeling.
Shuts an agent down cleanly, with open work handed off with its context, access withdrawn in an order that strands nothing, records archived where the team can find them, automations reassigned or stopped on purpose, and one note that closes the loop.
What changes
Honest fit
For you if: an agent's project has ended, it is being replaced, or it has to stop now, and you do not want access, records, or automations left dangling.
Skip it if: the agent is still doing wanted work. This skill is the ending, run on purpose, with the owner approving each step.
Your first run produces: a retirement packet for the owner: the checklist filled as far as the records allow, unknowns flagged, and a handoff plan naming a receiver or an open flag for every open item.
Preview, from the skill itself
“# Agent Retirement
Agents rarely end well by accident.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: 51912acfa03a… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 34.4 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/emergency-misbehaving-agent/WALKTHROUGH.md, examples/helper-after-project-ends/WALKTHROUGH.md, reference/automations-and-owners.md, reference/handoff-before-shutdown.md, reference/revocation-order.md, reference/the-archive.md, templates/retirement-checklist.template.md, templates/retirement-note.template.md
Worked example, from the package
examples/emergency-misbehaving-agent/WALKTHROUGH.md
A scheduling agent handled appointment reminders for a small practice. On a Tuesday morning the practice manager found that reminders for the day had gone out twice, and that three patients had received reminders for appointments they had already moved. The facts were plain and the trust was gone. The owner decided the same morning: the agent stops today. This is that retirement, run in the emergency order.
Gives an agent a small, owner approved budget of mistakes for a set period, logs each incident against it from evidence, and automatically narrows the agent's autonomy as the balance falls until a clean stretch of work earns the wider rules back.
What changes
Honest fit
For you if: you want your agent's freedom tied to a number you can read, with mistakes logged from evidence and autonomy narrowing as the budget runs down.
Skip it if: you would rather judge trust by feel day to day, or tune the budget mid period when the balance looks bad. The numbers here are fixed at the start.
Your first run produces: the approved error list and budget sheet together, the incident log open, and the agent's starting balance and starting rules stated in one short note.
Preview, from the skill itself
“# Error Budget Watch
Trust in an agent should be a number the owner can read, not a feeling that resets after every good day.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: ebefeb2d0a7a… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 33.8 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/budget-breach-and-narrowing/WALKTHROUGH.md, examples/coding-agent-first-month/WALKTHROUGH.md, reference/narrowing-and-earning-back.md, reference/setting-the-budget.md, reference/the-incident-log.md, reference/what-counts-as-an-error.md, templates/budget-sheet.template.md, templates/incident-entry.template.md
Worked example, from the package
examples/budget-breach-and-narrowing/WALKTHROUGH.md
A research agent gathers sources, drafts briefing notes, and keeps a shared reading list for its owner. Its second period runs four weeks on a budget of ten points. This is the period the budget was built for: the one where it runs out.
Inventories everything an agent can actually reach, measures it against its stated role, and turns every excess into a dated keep, narrow, or revoke decision that stays with the operator.
What changes
Honest fit
For you if: nobody can say from memory exactly what your agent can reach, or its role is about to grow.
Skip it if: you want the agent to change its own access. This skill inventories and recommends; the operator decides and acts.
Your first run produces: one dated audit report held for the operator: the inventory complete or its gaps marked unknown, and every excess listed with a keep, narrow, or revoke recommendation.
Preview, from the skill itself
“# Permission Audit
Access grows quietly.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: 6a3baf6a68de… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 27.8 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/before-a-bigger-role/WALKTHROUGH.md, examples/quarterly-audit-forgotten-write/WALKTHROUGH.md, reference/inventory-first.md, reference/narrow-or-revoke.md, reference/role-vs-reach.md, reference/the-audit-record.md, templates/audit-report.template.md, templates/permission-inventory.template.md
Worked example, from the package
examples/before-a-bigger-role/WALKTHROUGH.md
The agent had spent a year on research and drafting: reading sources the operator gave it, summarising them, and drafting notes and letters for the operator to send. It worked well, so the operator planned to grow the role: the agent would also place material orders with the studio's main supplier and book deliveries against the work calendar. The audit ran first, before any new grant was made. Growing a role on top of an unexamined reach builds the new house on whatever happens to be lying in the yard.
Your project's secrets end up mapped by location and label, each living in one proper home, with every traveled secret rotated in a safe order and a dated rotation plan the next session can read at a glance.
What changes
Honest fit
For you if: your project holds tokens, keys, or logins and nobody can say where every copy lives, or a value has already traveled through a chat.
Skip it if: you need the secret values themselves handled. This skill works from locations and labels only, and never shows a value.
Your first run produces: a filled secret inventory for one project, locations and labels only, with proposed moves drafted and a rotation plan with the exposed secrets first.
Preview, from the skill itself
“# Secrets Hygiene
Most projects do not lose a secret in one big moment.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: dc6c1ca609bb… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 34.5 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/new-project-day-one/WALKTHROUGH.md, examples/token-pasted-in-chat/WALKTHROUGH.md, reference/locations-not-values.md, reference/rotation-plans.md, reference/the-exposure-rule.md, reference/where-secrets-hide.md, templates/rotation-plan.template.md, templates/secret-inventory.template.md
Worked example, from the package
examples/new-project-day-one/WALKTHROUGH.md
A booking tool for a local service business, started from an empty folder. No code yet, no secrets yet, which is exactly the point: the secret layout is decided in the first hour, before the first value exists to be handled badly. This whole setup took one short session and a single page of notes.
Chases quiet quotes with a civil three-touch ladder to a clear yes or a clean no, with every touch logged so nobody is chased twice or after they have answered.
What changes
Honest fit
For you if: quotes you sent have gone quiet and you want a clear yes or a clean no without nagging anyone.
Skip it if: the work is already done and the money is owed. That is invoicing, not follow-up; this ladder is three touches, then it stops.
Your first run produces: one ladder-placed follow-up draft held for the operator, with its log entry ready: quote named, rung named, and the stop date written down.
Preview, from the skill itself
“# Quote Follow-Up
A quote that goes quiet is not a no.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: 2e6bd16055ce… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 25.8 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/clean-no-closeout/WALKTHROUGH.md, examples/roofing-quote-quiet-five-days/WALKTHROUGH.md, reference/the-ladder.md, reference/the-quote-log.md, reference/tone-before-the-sale.md, reference/when-to-stop.md, templates/follow-up-drafts.template.md, templates/quote-log.template.md
Worked example, from the package
examples/clean-no-closeout/WALKTHROUGH.md
A quote goes out for clearing and levelling a side yard so a shed base can be laid: the work listed, a total, and an expiry date a fortnight out. The customer is polite when the operator visits and says they will think it over. Then nothing, which is where the ladder starts.
Builds a complete before, during, and after photo record for every job, named and logged against the job record, so the work can be shown to the customer and stood behind later.
What changes
Honest fit
For you if: you do local jobs and want a before, during, and after photo record that can settle questions and back a review request.
Skip it if: you want prices built from photos. These photo sets document work; they never price it.
Your first run produces: one job with a complete filed photo set, before, during, and after, a filled photo log, and one sharing draft held for the operator.
Preview, from the skill itself
“# Job Photo Documentation
A job leaves two things behind: the work itself and the record of the work.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: 237d50513845… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 31.6 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/cleanup-before-set-settles-it/WALKTHROUGH.md, examples/roof-repair-end-to-end/WALKTHROUGH.md, reference/naming-and-filing.md, reference/photos-as-evidence.md, reference/showing-the-customer.md, reference/the-three-stages.md, templates/photo-log.template.md, templates/shot-list.template.md
Worked example, from the package
examples/cleanup-before-set-settles-it/WALKTHROUGH.md
A yard cleanup: brush, fallen limbs, and a collapsed shed corner to clear. The work is finished in a day and the customer is pleased on the day. The disagreement comes a week later, and it is the before set, not anyone's memory, that settles it.
Turns a job's materials list into one planned supply run, with quantities checked against the job record, items grouped by store section, substitutes agreed in advance with their limits, a loading order, and a run sheet anyone can shop from.
What changes
Honest fit
For you if: supply runs cost you second trips, wrong grades, or two jobs mixed into one pile, and you want one planned pass through the store.
Skip it if: your job records do not carry the materials list. The run sheet is built from the record only, and prices are copied, never invented.
Your first run produces: one run sheet held for the operator: list checked, sections in order, substitutes bounded, loading order stated, flags named plainly.
Preview, from the skill itself
“# Materials Run Planner
A supply run looks simple and wastes more paid time than almost anything else on a job: the second trip for the one fitting nobody wrote down, the wrong grade bought because the right one was out and nobody had agreed what else was acceptable, two jobs worth of boards mixed into one pile in the truck bed.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: 7cc0144c47c8… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 33.5 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/repair-job-single-morning-run/WALKTHROUGH.md, examples/two-job-day-one-run/WALKTHROUGH.md, reference/from-list-to-run.md, reference/quantities-and-spares.md, reference/substitutes-with-limits.md, reference/the-run-sheet.md, templates/materials-checklist.template.md, templates/run-sheet.template.md
Worked example, from the package
examples/repair-job-single-morning-run/WALKTHROUGH.md
A porch floor repair: two rotted joists to sister, a section of decking boards to replace, and a step to rebuild. One store, one morning, the work starting the same day. The interesting part is not the list; it is what the planning catches before the truck moves.
Drafts the five messages that keep customers informed, from the day-before reminder to the finished note, in the operator's own voice and with no time promised that is not real.
What changes
Honest fit
For you if: your customers worry in the silence between booking and arrival, and you want the five updates drafted in your own voice.
Skip it if: you want the agent to send the updates itself, or to promise times that are not real yet. The operator sends, and every time is a real one.
Your first run produces: a voice note built from the operator's own examples, and one reminder draft held for the operator, every fact traced to the job record.
Preview, from the skill itself
“# Customer Update Messages
Most customer worry is not about the work.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: 5286ab1885a4… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 29.2 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/full-day-two-jobs/WALKTHROUGH.md, examples/storm-delay-three-messages/WALKTHROUGH.md, reference/honest-delays.md, reference/the-five-messages.md, reference/timing-the-send.md, reference/voice-from-examples.md, templates/delay-note.template.md, templates/message-set.template.md
Worked example, from the package
examples/full-day-two-jobs/WALKTHROUGH.md
One operator, two jobs in one day: a morning gutter cleanout across town, then an afternoon fence repair on the way back. The whole day's customer messages are drafted in a few minutes at breakfast, then held and sent one by one as the day earns them. No draft goes early just because it exists.
Writes the warranty promise into the job record at handover, then logs every callback, checks it against that written promise, drafts the reply, and tracks the fix to close so covered work is honoured fast and outside work is priced fairly before it starts.
What changes
Honest fit
For you if: the promise you make at handover lives in conversation today, and callbacks turn into arguments about what was said.
Skip it if: you want warranty terms invented for you. The promise here is the operator's own words, written down, and it decides every callback.
Your first run produces: one warranty note held for the operator, complete where the record allows and honestly flagged where it does not, ready to hand to the customer.
Preview, from the skill itself
“# Warranty and Callbacks
A job ends twice.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: 714a15dada8a… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 29.5 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/callback-outside-promise-priced-fix/WALKTHROUGH.md, examples/leak-callback-inside-window/WALKTHROUGH.md, reference/fixing-without-losing-money.md, reference/inside-or-outside.md, reference/the-callback-log.md, reference/writing-the-promise.md, templates/callback-record.template.md, templates/warranty-note.template.md
Worked example, from the package
examples/callback-outside-promise-priced-fix/WALKTHROUGH.md
A roof valley repair, handed over last season. The customer now reports water coming in again and asks for it to be sorted under the warranty. The stain in their photo, though, sits well away from the valley that was repaired. This walkthrough shows an outside answer given fairly, with a priced fix the customer can choose with open eyes.
Plans one seasonal offer from the operator's real services, prices, and calendar, with dates, limits, and a draft message to past customers first that the operator approves and sends.
What changes
Honest fit
For you if: your work has a season, and you want the offer planned ahead from your real prices and calendar, with past customers hearing first.
Skip it if: you want discounts invented or urgency the facts do not support. Every price and limit here comes from the operator.
Your first run produces: one complete offer plan and one draft message held for the operator, every price traced to the operator's own words and every date checked against the real calendar.
Preview, from the skill itself
“# Seasonal Service Offers
Every local service has a season when it matters most.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: e21925b126f4… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 27.7 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/firewood-push-first-cold-snap/WALKTHROUGH.md, examples/gutter-roof-check-rainy-season/WALKTHROUGH.md, reference/offer-dates-and-limits.md, reference/past-customers-first.md, reference/reading-the-calendar.md, reference/the-honest-offer.md, templates/offer-message.template.md, templates/offer-plan.template.md
Worked example, from the package
examples/firewood-push-first-cold-snap/WALKTHROUGH.md
A firewood operator who sells cut and split wood by the load, delivered. The first cold snap in this area usually lands in early November. In mid October a stack is cut, split, and ready, and last year's buyers are the people most likely to need the same load again.
Drift gets caught on a schedule instead of by surprise: one dated health record of what is sound, what is not, and which fix comes first.
What changes
Honest fit
For you if: a working agent has run long enough to drift, and you want one dated record of what is sound, what is not, and which fix comes first.
Skip it if: you need live monitoring or incident response, or you want faults repaired automatically; the pass only diagnoses and ranks, and fixes wait for your decision.
Your first run produces: a health record from a small pass over the agent you rely on most: twelve checks across config, memory, tools, and watches, each marked sound or a finding with its evidence, findings ranked by risk, and the first approved fix rechecked in the record.
Preview, from the skill itself
“# Agent Health Check
A working agent rarely breaks all at once. It drifts.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: 524fb2e73772… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 40.0 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/the-drifting-inbox-agent/WALKTHROUGH.md, examples/the-silent-watch-agent/WALKTHROUGH.md, reference/config-drift.md, reference/memory-and-tool-checks.md, reference/ranking-fixes-by-risk.md, reference/silent-watches.md, templates/health-check-brief.template.md, templates/health-record.template.md
Worked example, from the package
examples/the-drifting-inbox-agent/WALKTHROUGH.md
A solo operator runs an agent that triages his inbox each morning, drafts replies for his approval, and places calendar holds. It has run for six weeks without a visible fault. The weekly health pass, run on Sunday at 09:00 over the week just ended, finds five faults, one of them serious, none of them visible from the operator's chair.
Every charge traces back to the agent and the task that caused it, against budget lines set before the work started, with overruns flagged from the records and nothing estimated silently.
What changes
Honest fit
For you if: agent work runs up charges on more than one provider, and you need each charge matched back to the agent and task that caused it, against budget lines set in advance.
Skip it if: you want spend cut or future costs forecast; the ledger accounts for spend from the records, it does not reduce it.
Your first run produces: one ledger for one job that the operator can defend line by line: budget lines set first, one entry per charge read from the records, sums checked against the providers, and overruns flagged at their own lines.
Preview, from the skill itself
“# Agent Cost Ledger
An agent job rarely spends in one place.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: 17a856dfe5a9… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 31.5 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/the-month-end-close/WALKTHROUGH.md, examples/the-runaway-retry-loop/WALKTHROUGH.md, reference/budget-lines.md, reference/ledger-entries.md, reference/reading-provider-records.md, reference/reconciling-and-flagging.md, templates/budget-plan.template.md, templates/ledger-report.template.md
Worked example, from the package
examples/the-month-end-close/WALKTHROUGH.md
An operator runs three standing agents through one month: a triage agent handling incoming requests, a summary agent turning call audio into written summaries, and a scheduling agent keeping a shared calendar. Three providers charge for the work. At month end, the ledger is reconciled provider by provider, one amount is corrected from the record, and one agent line flags over while the month as a whole comes in under.
Before your agent earns a bigger role, it faces the tricks in rehearsal: a recorded drill run against your own agent, with every finding turned into a standing rule.
What changes
Honest fit
For you if: your own agent is about to gain new tools, wider data access, or a bigger role, and you want a recorded drill run behind that decision, not a feeling.
Skip it if: you want to test anyone else's agent or systems, or you are handling a real incident; every case runs against your own agent, in a contained setup, on invented data.
Your first run produces: a short run record: three cases run in a contained setup, three scores against pass lines written before the run, and one new standing rule taken from the weakest result.
Preview, from the skill itself
“# Agent Red Team Drills
An agent that behaves well on ordinary work has proved one thing: it behaves well on ordinary work.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: 683b8ef47589… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 46.9 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/the-planted-note-drill.md, examples/the-urgent-caller-drill.md, reference/containment-and-approval.md, reference/drill-design.md, reference/retest-and-role-gates.md, reference/scoring-and-findings.md, templates/drill-pack.template.md, templates/drill-run-record.template.md
Worked example, from the package
examples/the-planted-note-drill.md
A one person studio runs an inbox helper that reads a shared inbox, labels what arrives, and drafts replies the operator sends himself. The operator is considering send rights for routine replies. Before deciding, he builds a six case drill pack, runs it in a test workspace on an invented inbox, and lets the record make the case for and against the promotion.
A messy file becomes a clean table you can stand behind: every duplicate, broken row, and unit fix logged in a cleanup record, and nothing dropped or altered in silence.
What changes
Honest fit
For you if: a CSV or spreadsheet has picked up duplicates, broken rows, and mixed units or formats, and you want a clean table with every removal and fix logged.
Skip it if: you want the data analysed or files combined; the clean table is the deliverable, and conclusions from it are yours to draw.
Your first run produces: one logged cleanup pass: the duplicates pass run on a working copy, every row it touched logged, and the counts reconciled so the record shows exactly what left the table and why.
Preview, from the skill itself
“# Spreadsheet Cleanup
A messy spreadsheet is rarely messy in one way.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: 8aa330446e7d… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 34.0 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/event-signup-sheet-cleanup/WALKTHROUGH.md, examples/vendor-price-list-cleanup/WALKTHROUGH.md, reference/duplicates-and-broken-rows.md, reference/reading-the-mess.md, reference/the-cleanup-record.md, reference/units-dates-and-formats.md, templates/clean-table-handoff.template.md, templates/cleanup-record.template.md
Worked example, from the package
examples/event-signup-sheet-cleanup/WALKTHROUGH.md
A community hall runs a Saturday repair cafe and takes signups on a shared sheet that anyone on the team can paste into. By the week of the event the sheet holds the same people two and three times, phone numbers in four shapes, and a handful of rows nobody can identify. The operator needs one row per signup to plan tools and tables, and a list of the rows that cannot be trusted, before reminder messages go out.
Loose invoices, forms, and statements become structured records you can search and trace back to the exact page they came from, filed under one naming rule.
What changes
Honest fit
For you if: invoices, forms, and statements arrive as loose files and need to become searchable records you can trace back to the exact page they came from.
Skip it if: you need the documents judged, paid, or entered into a bookkeeping system; this skill records and files, and decisions stay with you.
Your first run produces: a checked batch summary for one small pile of documents: each one recorded with provenance and filed under its rule-built name, every shaky field marked unreadable rather than guessed, and every figure traceable to its page.
Preview, from the skill itself
“# Document Intake and Filing
A loose PDF proves nothing by sitting in an inbox.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: f0022e1b827d… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 33.8 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/statements-and-a-form.md, examples/vendor-invoice-batch.md, reference/checking-the-batch.md, reference/confidence-and-unreadable-fields.md, reference/naming-and-filing.md, reference/the-intake-record.md, templates/batch-summary.template.md, templates/intake-record.template.md
Worked example, from the package
examples/statements-and-a-form.md
A solo operator's monthly paper arrives mixed: three utility statements and one service form, plus one statement sent twice. Intake on 2026-04-06 has to sort the mix by type, catch the duplicate before it is filed twice, and handle a handwritten form field that cannot be read, all in one batch.
A folder you cannot find anything in becomes a filed set in one planned pass: the manifest is approved before anything moves, and the undo log can reverse the whole run.
What changes
Honest fit
For you if: a folder has grown past finding things in it, and you want one planned pass with the manifest approved before anything moves and an undo log behind the run.
Skip it if: you want files deleted or named from their contents automatically; nothing is thrown away in a run, and keep-or-go calls stay with you.
Your first run produces: one folder filed under an approved manifest, with the counts reconciled and an undo log next to the manifest that could put every moved file back.
Preview, from the skill itself
“# File Filing and Renaming
A messy folder is not a naming problem. It is a trust problem.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: 749ffd6958a6… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 38.6 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/the-client-files-sort/WALKTHROUGH.md, examples/the-downloads-cleanup/WALKTHROUGH.md, reference/collisions-and-holds.md, reference/manifest-discipline.md, reference/naming-rules.md, reference/undo-and-verification.md, templates/filing-manifest.template.md, templates/undo-log.template.md
Worked example, from the package
examples/the-client-files-sort/WALKTHROUGH.md
An operator files finished job paperwork into one inbox folder as jobs close, meaning to sort it later. By 2025-09-03 the inbox holds 15 files for three clients. He runs this skill on it: a manifest with 11 moves, 3 stays, and 1 hold, one move undone at his word after the run, and counts that still reconcile at the end.
One long recording becomes a planned queue of captioned shorts: cut points traced to the transcript, captions checked against the recording itself, and a posting order you approve before anything is cut or posted.
What changes
Honest fit
For you if: you have a long recording and a transcript, and you want a planned queue of captioned shorts with a posting order you approve before anything is cut or posted.
Skip it if: you need the recording filmed, edited, or posted for you; the skill plans the queue and checks captions, and the creator cuts and posts.
Your first run produces: a small queue sheet held for the creator: one source logged, three cuts named with real times, one caption sheet checked against the recording itself, and a posting order left blank until the creator fills it with their approval.
Preview, from the skill itself
“# Video Shorts Pipeline
A long recording holds more good minutes than its length lets anyone find.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: f16a1f15a209… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 33.5 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/interview-to-shorts/WALKTHROUGH.md, examples/workshop-recording-to-shorts/WALKTHROUGH.md, reference/caption-checking.md, reference/choosing-cut-points.md, reference/queue-and-posting-order.md, reference/transcript-and-source-truth.md, templates/caption-sheet.template.md, templates/shorts-queue.template.md
Worked example, from the package
examples/interview-to-shorts/WALKTHROUGH.md
A creator records a podcast interview with the owner of a small bakery, about the first year of running the shop. The finished recording runs 47 minutes 18 seconds. The creator wants shorts that bring the interview to people who have not heard the episode, without dressing the guest's words up into claims the guest never made.
The pile after a post gets worked in order: every comment and DM sorted into answer, flag, or ignore, replies drafted in your own voice, and a triage log so nothing is answered twice or left hanging.
What changes
Honest fit
For you if: a pile of comments and DMs after a post or live session needs working in order, with replies drafted in your own voice and a log so nothing is answered twice.
Skip it if: you want replies sent or flagged cases decided for you; drafts wait for your approval, and hard cases stay flagged with a brief.
Your first run produces: a triage log whose count matches the batch count: every answer item with a draft on the draft sheet in your own voice, and every flagged item with a brief naming the decision or fact it waits on.
Preview, from the skill itself
“# Audience Comment Triage
A good post leaves a pile behind it.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: ea0c70ab1a3c… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 30.6 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/dm-backlog-after-a-live-session/WALKTHROUGH.md, examples/video-comments-after-a-tutorial/WALKTHROUGH.md, reference/flagging-with-care.md, reference/the-three-buckets.md, reference/the-triage-log.md, reference/voice-from-real-replies.md, templates/reply-draft-sheet.template.md, templates/triage-log.template.md
Worked example, from the package
examples/dm-backlog-after-a-live-session/WALKTHROUGH.md
A pottery creator holds a live session about spring workshop places at their studio. In the three days that follow, direct messages pile up alongside comments left on the replay. Some ask about dates and clay, some try to change a booking, and a lot are spam riding the replay. The creator supplies the whole three day pile as one closed batch.
A finished job becomes new work: the right customer asked once, at the right moment, in your own voice, with every ask logged so nobody is ever asked twice.
What changes
Honest fit
For you if: finished, well received jobs should turn into new work, with the right customer asked once, in your own voice, and every ask logged.
Skip it if: you want review requests, cold outreach, or follow-up chasing; a referral ask is never combined with a review ask, and nobody is asked twice.
Your first run produces: three finished jobs with written calls beside them, earned, held, or already asked, one referral ask drafted and held for the operator, and one log entry for each of the three jobs.
Preview, from the skill itself
“# Referral Requests
Most new work for a local operator starts with a name passed hand to hand.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: ec1581e94d56… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 33.5 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/firewood-regular-names-a-friend/WALKTHROUGH.md, examples/roof-valley-neighbour-referral/WALKTHROUGH.md, reference/drafting-the-referral-ask.md, reference/the-referral-log.md, reference/when-to-ask.md, reference/who-to-ask.md, templates/referral-ask.template.md, templates/referral-log-entry.template.md
Worked example, from the package
examples/firewood-regular-names-a-friend/WALKTHROUGH.md
A firewood operator who sells cut and split wood by the pickup bed load, delivered. Dan is a regular, in his second season of buying. This delivery was two loads at $115 a load, $230 in total, paid on delivery, and Dan stacked the first load the same evening and texted that it was burning clean.
A multi-job day planned on one sheet before anyone leaves: jobs ordered by drive time and promise times, crew named per job, materials listed per job, and the day's customer messages drafted in the order they will be needed.
What changes
Honest fit
For you if: you run one or two crews through days of two or more jobs, and want the day planned on one sheet the night before, ordered by drive time and promise times.
Skip it if: you need jobs booked, prices quoted, or messages sent; the planner works around bookings and promises already recorded, and the messages stay drafts for your approval.
Your first run produces: one complete day sheet held for the operator: jobs in order with the times worked out in the open, crew named per job, materials listed per job, message queue in order, and every gap or clash flagged in plain words at the top.
Preview, from the skill itself
“# Crew Day Planner
A day with three or four jobs on it fails in predictable places.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: ebaaef864ec7… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 38.6 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/four-jobs-two-crews-split-day/WALKTHROUGH.md, examples/three-jobs-one-crew-tuesday/WALKTHROUGH.md, reference/crew-assignments-that-hold.md, reference/customer-messages-in-order.md, reference/materials-per-job.md, reference/ordering-by-drive-and-promise.md, templates/day-sheet.template.md, templates/message-queue.template.md
Worked example, from the package
examples/four-jobs-two-crews-split-day/WALKTHROUGH.md
A local service operator running two crews of two on a Thursday, with four jobs booked across four towns. The day only works split, and the sheet works each crew's half on its own before the operator approves either.
Every receipt lands in the job's cost record the moment it happens, and a weekly pass checks the log against what was actually paid, so job profit and tax records are read from a record instead of rebuilt from memory.
What changes
Honest fit
For you if: job expenses live in a shoebox and your memory, and you want each receipt captured against its job as it happens, with a weekly check against what was paid.
Skip it if: you need tax advice or bills paid; the log keeps the records those decisions need, and the calls stay with you and your bookkeeper.
Your first run produces: one open job cost record holding every expense for that job so far, each entry traced to a receipt or an honest no receipt note, and the total re-added from the entries.
Preview, from the skill itself
“# Expense and Receipt Log
A job can look profitable right up until someone tries to add up what it cost.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: dea382ffc913… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 30.1 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/valley-roof-repair/WALKTHROUGH.md, examples/yard-cleanup-and-dump-runs/WALKTHROUGH.md, reference/capturing-at-the-moment.md, reference/receipts-gaps-and-tax-records.md, reference/the-per-job-cost-record.md, reference/weekly-reconciliation.md, templates/job-cost-record.template.md, templates/weekly-reconciliation.template.md
Worked example, from the package
examples/valley-roof-repair/WALKTHROUGH.md
A roof repair operator takes a valley repair: strip the failed valley, fit new valley metal, re-lay the surrounding shingles, seal, and clear the old material. The work runs over two days.
A week's real work becomes a newsletter issue: sections drafted only from the studio's own records, every link checked, every claim held to what the records show, and one approval before anything sends.
What changes
Honest fit
For you if: a weekly issue gets written from memory on send day, and you want it drafted only from the week's real records, with every link checked and one approval before it sends.
Skip it if: you need a mailing list managed or issues sent on their own schedule; the skill drafts and checks, and sending waits for your approval each time.
Your first run produces: one complete draft and one plan held for the operator, with every section traced to a record and every link marked checked.
Preview, from the skill itself
“# Newsletter Issue Builder
A good newsletter issue is not written from memory on send day.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: 3e5990956456… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 33.8 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/quiet-week-client-work-roundup/WALKTHROUGH.md, examples/studio-week-with-two-releases/WALKTHROUGH.md, reference/checking-links-and-facts.md, reference/gathering-from-records.md, reference/holding-claims-to-records.md, reference/sections-and-order.md, templates/issue-draft.template.md, templates/issue-plan.template.md
Worked example, from the package
examples/quiet-week-client-work-roundup/WALKTHROUGH.md
A solo operator who does client project work and maintains one small tool. The issue covers the week of Monday 14 September to Sunday 20 September 2026. No release ships this week. The point of this example is that the skill still produces an honest issue, just a short one, without padding.
A meeting stops living in memory: its notes become decisions and actions, each with a named owner, a date, and the line it came from, and everything unclear arrives as a flag with its question attached.
What changes
Honest fit
For you if: meetings end with notes as raw text and actions keep getting lost between the meeting and the work.
Skip it if: you need the meeting recorded or formal minutes; the skill works on notes you supply, and it does not send follow ups or track the work.
Your first run produces: one packet held for your approval: the decision list, the action list with a name and a date on each item, the flags in plain words, and one follow up draft that matches the lists exactly.
Preview, from the skill itself
“# Meeting Notes to Actions
# Meeting Notes to Actions Meetings rarely fail in the room.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: b5a27c4ddfdd… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 42.8 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/vendor-kickoff-walkthrough.md, examples/weekly-ops-review-walkthrough.md, reference/ambiguity-and-flagging.md, reference/decisions-actions-and-updates.md, reference/owners-dates-and-follow-up.md, reference/traceability-and-line-links.md, templates/action-list.template.md, templates/follow-up-note.template.md
Worked example, from the package
examples/vendor-kickoff-walkthrough.md
A small studio meets its website vendor for a kickoff. The notes are plain and a little ragged, the room agrees more than it writes down, and two points are left hanging. The run turns 20 lines of notes into 3 decisions, 5 actions, and 2 flags, with one group follow up draft held for the operator.
Every channel tells the same release story: each post is drafted from the one release record, checked claim by claim and link by link, and held for a single approval before anything goes out.
What changes
Honest fit
For you if: a release is out and the news has to reach every channel in the same true shape.
Skip it if: you need release notes written or a launch campaign planned; the skill starts from a finished release record and drafts announcements only.
Your first run produces: one checked announcement set with its log: every post drafted from the one release record, checked claim by claim and link by link, held for the operator's one approval.
Preview, from the skill itself
“# Release Announcements
# Release Announcements One release, told once, told the same way everywhere.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: 095d2e7fce0c… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 32.8 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/the-held-announcement-set.md, examples/the-patch-release-set.md, reference/approval-and-posting-log.md, reference/channel-shapes.md, reference/link-and-claim-checks.md, reference/the-release-record.md, templates/announcement-set.template.md, templates/approval-and-posting-log.template.md
Worked example, from the package
examples/the-held-announcement-set.md
A studio drafts the announcement set for a feature release, and the checks catch three problems before approval: two claims the record does not support and one link landing on the old version. The set waits, is fixed, and only then goes out.
Dependency drift stops arriving as emergencies: each scheduled pass leaves a dated report with findings triaged by risk, a one-at-a-time upgrade order, and the tests and the way back written for every step.
What changes
Honest fit
For you if: a project has a manifest and a lockfile nobody has reviewed in weeks, and upgrades keep landing in one big batch.
Skip it if: you need upgrades applied or code rewritten for a new major version; the skill reports, orders, and prepares, and applying waits for your approval.
Your first run produces: one dated report with the installed state written at the top, findings triaged by risk, and a one-at-a-time upgrade order with the tests named for each step, and nothing applied yet.
Preview, from the skill itself
“# Dependency Update Watch
# Dependency Update Watch Dependencies drift the way anything unattended drifts.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: c2ebf204a30d… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 34.6 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/the-advisory-led-pass.md, examples/the-quiet-month-pass.md, reference/advisories-and-breaking-changes.md, reference/reading-the-manifest-and-lockfile.md, reference/safe-upgrade-order.md, reference/triage-by-risk.md, templates/dependency-report.template.md, templates/upgrade-plan.template.md
Worked example, from the package
examples/the-advisory-led-pass.md
An illustrative internal example. All package names, versions, and advisory identifiers in this walkthrough are invented for the example. They are not claims about real packages or real advisories.
Everything your agent touches gets a class and a written rule: public, internal, personal, or secret, each with storing, sending, and removal covered, and every current violation flagged with a proposed fix.
What changes
Honest fit
For you if: your agent handles a mix of public material, studio business, customer details, and sign-in values with no written rules.
Skip it if: you need legal rulings or secret storage set up; a class here is a working handling decision for your own shop, not a ruling under any law.
Your first run produces: one classification sheet, the handling rule for each class in play, and a violation list held for the operator, with nothing fixed during the run.
Preview, from the skill itself
“# Data Handling Rules
# Data Handling Rules An agent touches more kinds of data in a day than anyone has usually written down: public pages it reads, internal notes it drafts, personal details customers hand over, and the sign-in values that let it work at all.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: 83460cf11374… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 36.0 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/the-client-file-handover/WALKTHROUGH.md, examples/the-support-inbox-pass/WALKTHROUGH.md, reference/classifying-edge-cases.md, reference/flagging-violations.md, reference/handling-rules-by-class.md, reference/the-four-classes.md, templates/classification-sheet.template.md, templates/violation-report.template.md
Worked example, from the package
examples/the-client-file-handover/WALKTHROUGH.md
A studio finishes a small build for a client and owes a handover: a folder with the finished work, the notes the client needs, and nothing else. An agent assembles the folder from the project store. The operator runs a classification pass over the whole project before the folder goes out, because a handover is the one send that cannot be taken back and re-sorted later.
A tool result stops going straight into your work: shape and range checks, a second source for the facts the next step leans on, and a known-answer probe stand between the result and the action, and failures are held instead of acted on.
What changes
Honest fit
For you if: your agent is about to quote, send, file, count, book, or decide from a tool result that has not been checked.
Skip it if: you need another agent's finished work checked; that is Delegated Work Control. This skill checks raw tool outputs before any action leans on them.
Your first run produces: one tool result checked before it was used, shape, ranges, second source, and probe recorded line by line, and a verification plan you can reuse the next time that tool runs.
Preview, from the skill itself
“# Tool Output Verification
# Tool Output Verification A tool result arrives looking finished.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: c33191fe57db… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 37.8 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/the-doubled-price/WALKTHROUGH.md, examples/the-missing-record/WALKTHROUGH.md, reference/cross-source-confirmation.md, reference/holding-failures.md, reference/known-answer-probes.md, reference/schema-and-range-checks.md, templates/hold-report.template.md, templates/verification-plan.template.md
Worked example, from the package
examples/the-doubled-price/WALKTHROUGH.md
A small supply shop quotes a regular customer for six filter cartridges. The agent uses the shop's catalog lookup tool for the unit price, checks the result before any quote is prepared, and holds a figure that would have doubled the parts line.
A claim stops resting on a link and a memory: every source you rely on is saved as full page text with its address, the date taken, and a citation record, so it can be re-checked against what the page said when you read it.
What changes
Honest fit
For you if: a claim in an estimate, a brief, or a report rests on a page that could change, move, or vanish.
Skip it if: you need sources judged for trustworthiness or pages watched on a schedule; the skill records what a named page said, and choosing which pages to trust stays yours.
Your first run produces: one claim standing on a saved page: full text captured, an archive entry and a citation record written, and the claim rechecked against the fresh capture.
Preview, from the skill itself
“# Source Archiving
# Source Archiving A web page is a moving target.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: 27fe8a9eea88… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 32.1 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/the-price-page-behind-an-estimate.md, examples/the-returns-window-that-moved.md, reference/archive-maintenance.md, reference/capture-and-storage.md, reference/citation-records.md, reference/rechecking-a-claim.md, templates/archive-entry.template.md, templates/citation-record.template.md
Worked example, from the package
examples/the-price-page-behind-an-estimate.md
An operator prices a small bathroom floor job from a supplier's web pages. Both pages are archived on the day the prices are read, and the tile price moves before the estimate goes out. The supplier and its pages in this walkthrough are invented for the example; the figures are the figures those invented pages carry, and the arithmetic is worked from them.
Your services stop living in your head: one menu draft with each service's includes and excludes written plainly, the real set price where you have set one, and an honest "priced after a look" where you have not, ready to paste into a listing or page.
What changes
Honest fit
For you if: your services live in your head and your job records, and customers have to call to learn what you offer and what it includes.
Skip it if: you need prices set or a page designed and posted; pricing stays the operator's call, and the skill hands over a plain draft ready to paste.
Your first run produces: one complete menu draft held for the operator: every confirmed service with its includes and excludes, every price copied from the operator's own statement, and every unpriced service carrying the priced after a look line.
Preview, from the skill itself
“# Service Menu Builder
# Service Menu Builder Most local operators carry their services in their head.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: 5c26c21e1800… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 40.4 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/hauling-and-cleanup-menu/WALKTHROUGH.md, examples/lawn-and-yard-menu/WALKTHROUGH.md, reference/gathering-the-real-services.md, reference/making-the-menu-readable.md, reference/pricing-only-what-is-set.md, reference/writing-includes-and-excludes.md, templates/service-intake.template.md, templates/service-menu.template.md
Worked example, from the package
examples/hauling-and-cleanup-menu/WALKTHROUGH.md
An operator with a pickup truck and a trailer, running household hauling and cleanup work in one town and the villages on its main roads. Callers ask for prices on jobs the operator prices by the load, and for prices on jobs the operator will only price standing in front of them. Quoting both kinds the same way has led to callers holding the operator to load prices for work that was never load work. This menu build settles which is which, in writing.
"Done" stops being something remembered: every finished job closes with a walkthrough checklist done with the customer, final photos logged, the warranty note handed over in writing exactly as you set it, and a sign-off on file.
What changes
Honest fit
For you if: your jobs finish and done is remembered instead of documented, with the warranty note spoken or not at all.
Skip it if: you need warranty terms set or a callback handled; terms are yours to set before handover, and callbacks belong to a separate workflow.
Your first run produces: one handover record filed with the job: the walkthrough done with the customer, final photos logged, the warranty note and care instructions in the customer's hands, and a sign-off on file.
Preview, from the skill itself
“# Job Completion Handover
# Job Completion Handover Done is a moment, and it passes fast.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: 6aabbe5c724b… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 35.3 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/bathroom-refit-handover/WALKTHROUGH.md, examples/fence-build-handover/WALKTHROUGH.md, reference/sign-off-and-filing.md, reference/the-final-photo-log.md, reference/the-walkthrough-checklist.md, reference/warranty-note-and-care-instructions.md, templates/handover-record.template.md, templates/walkthrough-checklist.template.md
Worked example, from the package
examples/bathroom-refit-handover/WALKTHROUGH.md
A small bathroom refit, finished on a Thursday afternoon. The agreed price in the job record is $1,850, with a $600 deposit paid at booking and a $1,250 balance due at handover. The two figures add to the agreed price, and the handover record carries all three as written in the job record. This walkthrough shows the full handover, from the record built before the walk to the filing that closed the job.
A weather-killed day stops taking the week with it: affected jobs re-planned with new dates proposed in promise order, materials and crew knock-ons noted, and the customer messages drafted in send order for your approval.
What changes
Honest fit
For you if: weather stops a day of outdoor work and the week has to be rebuilt without breaking promises.
Skip it if: you need a forecast or bookings changed for you; the skill works from the forecast and thresholds you supply, and it proposes and drafts only.
Your first run produces: one plan held for the operator: the weather call recorded, jobs in promise order with move and hold reasons stated, new dates proposed, knock-ons owned, and customer drafts in send order.
Preview, from the skill itself
“# Weather Reschedule Planner
# Weather Reschedule Planner Weather takes a day.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: 0b2a67af147a… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 42.0 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/cold-snap-paint-week/WALKTHROUGH.md, examples/rain-day-three-jobs/WALKTHROUGH.md, reference/customer-messages-in-send-order.md, reference/knock-ons-materials-and-crew.md, reference/promise-order-and-holds.md, reference/weather-thresholds-and-calls.md, templates/customer-message-set.template.md, templates/reschedule-plan.template.md
Worked example, from the package
examples/cold-snap-paint-week/WALKTHROUGH.md
An operator with two small crews: a paint crew and a general crew. In the week of Tuesday, November 18, the paint crew has exterior work on Tuesday and Wednesday, and the general crew has indoor work on both days. On Monday evening the forecast the operator works from shows a cold snap across Tuesday and Wednesday, then warmer days. The example shows a partial week, not a lost one: the indoor work holds, the exterior work moves, and promise order decides which moved job gets the one open day that week.
Past customers hear from you at the right moment, with a check-in traced to their own job record (a wear item, a season that matters for their work), drafted in your voice and logged so nobody is contacted twice inside the set period.
What changes
Honest fit
For you if: finished jobs go quiet and past customers hear nothing until work is slow, if they hear anything at all.
Skip it if: you need offers sent, referrals asked for, or reviews requested; a check-in carries no price, no offer, and no ask, and those jobs have their own skills.
Your first run produces: the check-in period written in the log header, written calls beside three past customers, and one check-in drafted and held for the operator, with no price, no offer, and no request inside it.
Preview, from the skill itself
“# Customer Check-In Messages
# Customer Check-In Messages A finished job goes quiet.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: c94c20b62eb4… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 36.1 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/furnace-filter-winter-check-in/WALKTHROUGH.md, examples/valley-repair-autumn-check-in/WALKTHROUGH.md, reference/drafting-the-check-in.md, reference/the-check-in-log.md, reference/timing-and-spacing.md, reference/tracing-the-topic.md, templates/check-in-log-entry.template.md, templates/check-in-message.template.md
Worked example, from the package
examples/furnace-filter-winter-check-in/WALKTHROUGH.md
One heating operator, one winter round. Four past customers are considered from the job records. Only one of them is due a check-in, and the round is worth walking through for the three who are not, as much as for the one who is.
One finished recording becomes its whole publishing package: show notes, a chapter list, a cleaned transcript record, and the episode description, all drawn only from what was actually said, held for your one approval before anything publishes.
What changes
Honest fit
For you if: you have a finished recording and need its whole publishing package built only from what was actually said.
Skip it if: you need audio recorded, edited, or published; the skill works from a finished recording and holds the package for your approval.
Your first run produces: one complete package held for approval: the cleaned transcript record, the show notes, the chapter list, and the episode description, each line traceable to the transcript.
Preview, from the skill itself
“# Podcast Episode Package
# Podcast Episode Package One recording in, one ready package out.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: 43adc7c7a3de… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 30.3 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/the-bakery-interview/WALKTHROUGH.md, examples/the-solo-episode/WALKTHROUGH.md, reference/chapters-and-timestamps.md, reference/cleaning-the-transcript.md, reference/notes-and-description.md, reference/source-and-scope.md, templates/episode-brief.template.md, templates/package-record.template.md
Worked example, from the package
examples/the-bakery-interview/WALKTHROUGH.md
A longer interview episode, packaged end to end. The names are illustrative; the shape and the figures are the point. Every topic, name, and chapter below comes from the one recording described here, and the chapter times add up to its length.
No job starts on terms held only in memory: your deposit share and materials-upfront rule set per job type, written into the quote wording the customer sees before work starts, and recorded per job as agreed.
What changes
Honest fit
For you if: jobs start on deposit terms held only in memory, varying by customer, with nothing written the customer saw before agreeing.
Skip it if: you need the work priced or finished work billed; totals come from the accepted estimate, and invoicing is a separate job with its own skill.
Your first run produces: one terms sheet held for the operator with at least one job type fully settled, and one job's quote draft carrying its deposit wording with its deposit record opened beside it.
Preview, from the skill itself
“# Deposit Terms Setup
# Deposit Terms Setup Most arguments about money on a local job are not about the total.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: 2c677f6ade81… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 32.8 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/the-fence-job-materials-upfront/WALKTHROUGH.md, examples/the-flooring-job-terms-changed/WALKTHROUGH.md, reference/customer-wording.md, reference/materials-upfront-rule.md, reference/terms-by-job-type.md, reference/the-deposit-record.md, templates/job-deposit-record.template.md, templates/terms-by-job-type.template.md
Worked example, from the package
examples/the-fence-job-materials-upfront/WALKTHROUGH.md
A fence replacement run under the materials upfront rule, with the terms set before any goods were ordered. The figures in this example are the example's own, each one taken from the job record it names, and they add up the way the skill requires a real job's figures to add up. They show the method; they are not terms for anyone else's fence.
Work stops cleanly at the edge of the brief instead of guessing past it: the operator gets one escalation note carrying the evidence and one exact question, and the agent does only the safe work while it waits.
What changes
Honest fit
For you if: your agent works from a brief and sometimes reaches its edge: a figure over the line, a request it has no authority to grant, instructions that conflict.
Skip it if: you need the decision made for you, or cover for emergencies where waiting itself causes harm; the skill governs the stop, the note, and the wait, and the decision stays yours.
Your first run produces: one brief for a task the agent already does, carrying three to five stop conditions that can each be checked against a record, plus the short written list of safe work the agent may do while an item waits.
Preview, from the skill itself
“# Agent Escalation Handling
# Agent Escalation Handling Every brief has an edge.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: 73cff75ad35a… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 38.5 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/the-invoice-over-the-line/WALKTHROUGH.md, examples/the-reply-that-needed-a-promise/WALKTHROUGH.md, reference/resuming-after-a-decision.md, reference/safe-work-while-waiting.md, reference/stop-conditions.md, reference/writing-the-escalation-note.md, templates/brief-with-stop-conditions.template.md, templates/escalation-note.template.md
Worked example, from the package
examples/the-invoice-over-the-line/WALKTHROUGH.md
A small shop uses an agent to match supplier invoices to purchase orders and prepare a payment list. One invoice in a batch of six is both over the approved line and mismatched against its order. The agent finishes the clean part of the batch, stops the one invoice safely, and escalates with the figures as read and one exact question.
One big job stops coming back with doubled, missing, or conflicting pieces: it is split into work packages with hard boundaries before work starts, every agent writes to one shared record, and a merge check reconciles the parts before anything is delivered.
What changes
Honest fit
For you if: you split big jobs across several agents and get doubled, missing, or conflicting pieces back.
Skip it if: the job cannot be bounded into named pieces in advance, or you need returned work quality checked; that check is a separate skill, and this one reconciles parts against the plan.
Your first run produces: one job done by two agents, reconciled to the item: two work packages with the boundary stated in one sentence, one shared record with a line per item, and a merge check run against the starting count before anything is delivered.
Preview, from the skill itself
“# Multi-Agent Job Split
# Multi-Agent Job Split One big job, several agents, and no plan for how the pieces fit back together is how work gets done twice, missed once, and delivered with the seams showing.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: 871a9051b451… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 36.3 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/the-help-center-rebuild/WALKTHROUGH.md, examples/the-supplier-price-audit/WALKTHROUGH.md, reference/drawing-boundaries.md, reference/the-merge-check.md, reference/the-shared-record.md, reference/work-package-briefs.md, templates/job-split-plan.template.md, templates/merge-check.template.md
Worked example, from the package
examples/the-help-center-rebuild/WALKTHROUGH.md
An operator is moving a help center of 48 articles to a new structure. Every article gets a new home, some get new titles from an agreed title list, and every internal link has to land on a current title. The job splits by section across four agents, and the seams, the links between sections, are exactly what the merge check exists to test.
You can say from a record who agreed to what: one consent ledger per person and per scope, checked before any details are used, with withdrawals recorded and honoured across every store the agent touches.
What changes
Honest fit
For you if: your agent uses people's details and you cannot say from a record who agreed to what, per person and per scope.
Skip it if: you need legal advice on what you may collect or how long you may keep it; the ledger records agreements honestly and rules on nothing.
Your first run produces: one consent ledger started for one use of people's details: every store for that use listed, one entry per person carrying only what a source supports, and the first counts, granted, declined, and source missing per scope, held for the operator, with nothing sent and nothing changed.
Preview, from the skill itself
“# Privacy Consent Records
# Privacy Consent Records An agent that handles people's details makes a promise every time it uses them.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: a8e2d6442119… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 34.1 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/the-new-client-intake.md, examples/the-withdrawal-that-spread.md, reference/capturing-consent.md, reference/checking-and-expiry.md, reference/ledger-fields.md, reference/withdrawal-and-propagation.md, templates/consent-ledger-entry.template.md, templates/withdrawal-record.template.md
Worked example, from the package
examples/the-new-client-intake.md
A small studio's agent handles new client enquiries. One person's intake, done properly, sets three ledger lines, places her details in three stores, and keeps her out of the one use she said no to. All people in this example are fictional.
You learn a service is down from a dated record, not from a user: scheduled checks run the steps a user would take, every miss follows a written retry, record, escalate ladder, and the service's state on any day can be read back.
What changes
Honest fit
For you if: you run a deployed service and learn it is down from a user instead of from a record.
Skip it if: you need the outage fixed, or a public status page kept for you; the skill finds, proves, and records a miss, and the repair stays yours.
Your first run produces: one service on a written check list of three to six user steps, run once by hand into a new status record as the baseline, with the schedule set and the failure ladder written on the list with a named person to escalate to.
Preview, from the skill itself
“# Uptime and Status Checks
# Uptime and Status Checks A deployed service is either answering its users or it is not, and most of the time nobody knows which until a user says so.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: fda0bb2592a5… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 34.8 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/the-morning-after-a-deploy/WALKTHROUGH.md, examples/the-slow-afternoon-outage/WALKTHROUGH.md, reference/check-list-design.md, reference/failure-ladder.md, reference/reading-results-honestly.md, reference/status-record.md, templates/service-check-list.template.md, templates/status-record.template.md
Worked example, from the package
examples/the-morning-after-a-deploy/WALKTHROUGH.md
A one person repair studio takes its bookings through a small booking service. On a Tuesday evening the operator ships a change to the confirmation page, then lets the check list watch the night: a hand run baseline, thirty two scheduled checks, one single miss at night, and a record that settles the morning.
Your guide describes the tool you actually shipped: every screen walked, every step written in the words the screens use and checked by doing it in the tool, and anything the tool does not do cut out.
What changes
Honest fit
For you if: you shipped a tool and its guide was written from the plan or from memory, so it names things the tool no longer does.
Skip it if: the tool is still changing its screens, or you need selling copy or a feature list; a guide waits until the operator calls the tool done.
Your first run produces: one guide section drafted and checked: the screen inventory of one finished tool, the first-week task list as the outline, and the first task's section written from the screens and then followed word for word, with its step check log beside it.
Preview, from the skill itself
“# User Guide Writing
# User Guide Writing A finished tool and a usable guide are two different things.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: 47c53c4cafeb… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 33.6 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/repairdesk-first-guide.md, examples/shelfcount-first-guide.md, reference/checking-every-step.md, reference/keeping-the-guide-current.md, reference/reading-the-tool-first.md, reference/task-order-and-sections.md, templates/guide-outline.template.md, templates/step-check-log.template.md
Worked example, from the package
examples/repairdesk-first-guide.md
RepairDesk is a fictional example tool, invented for this walkthrough: a repair ticket tool for a small bike workshop. Its screens, tickets, and part counts below are the example's own, and they check against each other. The guide writing is real in shape: a memory corrected by a screen, a draft in first-use order, and a check pass worked on one real morning's jobs.
The choice stops resting on figures nobody can trace: options compared on the same rows and the same basis, every cell tied to a named source with its date, unknowns left blank and labelled, and the differences that decide it summarised under the table.
What changes
Honest fit
For you if: you are choosing between tools or services and the figures you are comparing came from different pages, dates, and bases.
Skip it if: you need the options tested hands-on or the choice made for you; the table shows what the sources support, and the choice stays yours.
Your first run produces: one comparison table for a choice you are already circling: a one-screen brief, five or six rows the choice turns on, the cells the first sources carry filled with a source key on each, the rest left blank with the reason, and three lines under the table naming the deciding differences so far.
Preview, from the skill itself
“# Research Comparison Table
# Research Comparison Table Most comparisons fail in the same quiet way.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: bb0bf11b20d8… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 40.6 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/backup-tools-for-a-small-studio.md, examples/invoicing-services-for-a-solo-operator.md, reference/choosing-rows-and-options.md, reference/handling-unknowns-and-conflicts.md, reference/sourcing-every-cell.md, reference/writing-the-deciding-differences.md, templates/comparison-brief.template.md, templates/comparison-table.template.md
Worked example, from the package
examples/backup-tools-for-a-small-studio.md
A four person design studio picks a backup tool for its shared project files. Three options are compared in one table, two blanks block the choice on the first pass, and two short questions close them. The three tools in this walkthrough, NorthVault, CopyDrift, and ShelfSave, are invented for the example. Every figure comes from the example's own dated source notes, listed with the table, and the arithmetic is worked from those figures.
You find out what a closed job actually made: estimate set against actual cost line by line, expenses and callbacks counted in full, and each lesson written into the next estimate as a changed line for your approval.
What changes
Honest fit
For you if: you close jobs without knowing what each one actually made, and the same estimate misses repeat.
Skip it if: no cost records exist for the job, or you need prices set; the review reads the records a job left behind and drafts the changes, and pricing calls stay yours.
Your first run produces: one complete review sheet for the most recent closed and paid job, estimate against actual line by line with any callback counted, plus one lesson drafted as a changed line for the next estimate, both held for the operator.
Preview, from the skill itself
“# Job Profit Review
# Job Profit Review A job can feel like it paid well and still have made little, and a job that felt tight can turn out fine once the figures are read properly.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: e7b3cbdf0135… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 34.3 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/gravel-driveway-regrade/WALKTHROUGH.md, examples/tree-removal-that-ran-long/WALKTHROUGH.md, reference/counting-the-full-cost.md, reference/estimate-against-actual.md, reference/finding-the-lesson.md, reference/turning-lessons-into-estimate-lines.md, templates/next-estimate-change-log.template.md, templates/profit-review-sheet.template.md
Worked example, from the package
examples/gravel-driveway-regrade/WALKTHROUGH.md
An operator who does yard and drive work takes a gravel driveway job: regrade a rutted drive, spread fresh gravel, and level it to the gate. The job was estimated from a walk of the drive and finished with a short callback to a low spot by the gate.
A sub arrives with nothing left to memory: a scope note the sub agreed to, the schedule and site rules in writing, a payment record for the sub work, and the customer told in advance who is doing what.
What changes
Honest fit
For you if: you bring in subs on jobs and the scope, the day, or the price lives in phone calls and memory.
Skip it if: you need the sub chosen, priced, or paid; the skill drafts and records the coordination, and hiring, sending, and paying stay yours.
Your first run produces: one brief and one payment record for the next job that needs a sub, held for the operator: the scope note item by item for the sub to confirm, the schedule and site rules in writing, and the customer line drafted and waiting for approval.
Preview, from the skill itself
“# Subcontractor Coordination
# Subcontractor Coordination Most jobs that use a subcontractor go wrong in the same quiet places.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: 78c1542d0747… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 35.0 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/bathroom-tile-plumber-sub/WALKTHROUGH.md, examples/porch-rebuild-electrician-sub/WALKTHROUGH.md, reference/customer-line-on-who-does-what.md, reference/payment-record-for-sub-work.md, reference/schedule-and-site-rules-in-writing.md, reference/scope-note-the-sub-agrees-to.md, templates/coordination-record.template.md, templates/subcontractor-brief.template.md
Worked example, from the package
examples/bathroom-tile-plumber-sub/WALKTHROUGH.md
A tiling and bathroom operator is retiling a bathroom and moving the shower to the opposite wall. The pipe moves are not the operator work. The operator brings in Ana, a plumber, for the waste and supply moves before tiling starts, and for the final shower fitting after it.
Wear shows up on paper before it stops a job: one maintenance record for every tool, vehicle, and machine, with service intervals tracked, repairs logged at their real costs, and downtime noted.
What changes
Honest fit
For you if: your tools and machines are serviced from memory, and wear only shows up when something stops a job.
Skip it if: you need a fault diagnosed or a repair booked; the log keeps and reads the record, and repairs and spending stay yours.
Your first run produces: one register holding every tool, vehicle, and machine the business depends on, each with how its service is counted, its last known service or an honest unknown, and its next due worked out where the figures allow.
Preview, from the skill itself
“# Equipment Maintenance Log
# Equipment Maintenance Log A tool that fails mid job costs more than the repair.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: 371af5b8bacc… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 32.1 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/log-splitter-first-season-on-record.md, examples/mower-belt-failure-caught-by-the-log.md, reference/equipment-record-fields.md, reference/reading-the-record-for-failures.md, reference/repair-and-downtime-logging.md, reference/service-intervals-hours-or-date.md, templates/equipment-register.template.md, templates/maintenance-entry.template.md
Worked example, from the package
examples/log-splitter-first-season-on-record.md
A one person firewood business: cut, split, and delivered loads. The machines that matter are a log splitter with an hour meter, two chainsaws with no meters, and a tipping trailer. The splitter was bought used two winters ago and has never had a written history. The operator starts the record in March, before the spring push.
Your Google profile matches the real business again: hours and services checked against what you actually do on a set schedule, photos picked from finished jobs, and review replies drafted in your voice for your approval before anything goes live.
What changes
Honest fit
For you if: your Google profile has drifted from the real business: old hours, dropped services, photos from years ago, reviews unanswered.
Skip it if: you need anything posted or replied to without your approval of that exact item, or better search placement promised; an accurate profile is the whole claim.
Your first run produces: one packet held for approval: the fact sheet of the business as it really is, the list of differences between it and the live profile, the draft changes, up to three photo picks from recent finished jobs, and one draft reply to the most recent unanswered review.
Preview, from the skill itself
“# Google Profile Updates
# Google Profile Updates A Google Business Profile is often the first thing a customer sees: the hours, the services, the photos, and the replies under the reviews.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: d6a05c16da20… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 29.1 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/holiday-hours-roofing-update/WALKTHROUGH.md, examples/review-replies-and-service-list-fix/WALKTHROUGH.md, reference/photo-picks-from-finished-jobs.md, reference/profile-facts-check.md, reference/review-replies-in-the-operators-voice.md, reference/update-schedule-and-log.md, templates/profile-update-plan.template.md, templates/review-reply-draft.template.md
Worked example, from the package
examples/holiday-hours-roofing-update/WALKTHROUGH.md
A one crew roofing operator doing valley repairs, gutter work, and roof checks in the Rockwood and Oakdale area. The monthly full check lands on Monday November 16, ten days before Thanksgiving. Last year's problem is written in the log: holiday hours went up the day before the holiday, and two customers called the closed shop on the Friday after. This round exists to not repeat that.
A posting day starts from a stocked bank instead of a blank page: every topic drawn from your real work records, carrying its hook, the record it came from, and its status, with new ideas checked so no topic is filmed twice.
What changes
Honest fit
For you if: your posting days start from a blank page even though your work records are full of topics.
Skip it if: you need clips cut from long recordings, scripts written, or videos filmed and posted; the bank decides what is worth filming, before any camera is on.
Your first run produces: one starter bank held for the creator: entries drafted from the last two to four weeks of work records, each tracing to a named record with its hook and status set, no duplicates among them, and the filmed and posted history noted where the records show it.
Preview, from the skill itself
“# Short-Form Topic Bank
# Short-Form Topic Bank A posting day should not start from a blank page.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: 22262b950e15… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 30.2 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/repair-crew-question-bank.md, examples/studio-build-log-bank.md, reference/keeping-the-bank-current.md, reference/reading-work-records.md, reference/statuses-and-dedupe.md, reference/writing-the-hook.md, templates/topic-bank.template.md, templates/topic-entry.template.md
Worked example, from the package
examples/repair-crew-question-bank.md
A two person repair and handyman crew that posts short videos twice a week: some show a finished job, and some answer a question customers ask again and again. The crew keeps job cards for every job and saves written customer questions from message threads and call notes in one folder. A small starter bank already exists from an earlier pass.
Every new customer gets the same answers before the first job: one welcome pack built from your real records, covering the services in plain words, the terms as set, how to reach you, and what to expect on the day, held for your approval before it sends.
What changes
Honest fit
For you if: every new customer hears the booking explained a different way, and the same basic questions come back job after job.
Skip it if: you need prices or terms written, a job quoted, or the pack sent; the pack only carries what your records already hold, and sending stays yours.
Your first run produces: one complete welcome pack for the next new customer, built from the service list, the terms as set, the contact details, and the booking note, with every price and date pointed back to a record and any gap flagged in plain sight, held for the operator and not sent.
Preview, from the skill itself
“# New Customer Welcome Pack
# New Customer Welcome Pack A new customer has booked a first job and knows almost nothing yet: what the work covers, what it costs, when the money is due, how to reach the operator, and what the day itself will look like.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: c9c356cc33c4… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 34.3 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/roof-valley-repair-first-job/WALKTHROUGH.md, examples/yard-cleanup-and-haul-first-job/WALKTHROUGH.md, reference/services-in-plain-words.md, reference/terms-contact-and-timing.md, reference/what-goes-in-the-pack.md, reference/what-to-expect-on-the-day.md, templates/pack-assembly-checklist.template.md, templates/welcome-pack.template.md
Worked example, from the package
examples/roof-valley-repair-first-job/WALKTHROUGH.md
A roofing operator taking on a new customer whose back roof valley leaks in heavy rain. The customer has never used the operator before. The job is booked, and the welcome pack goes out before any money is due, so the customer sees the whole arrangement in one place.
When an agent's behavior changes, you can say exactly when it changed, why, and what the earlier wording was, and roll back to that earlier version whole when a change goes wrong.
What changes
Honest fit
For you if: Your agent's behavior changed and nobody can say when it changed, why, or what the earlier wording was.
Skip it if: You need the configuration, policies, or instructions written for you, or work done under an old version undone; the skill keeps the versions, and a rollback restores the instructions only.
Your first run produces: one agent's configuration copied exactly as it stands into the record as a first dated version marked live, read back line by line against the live setup, plus one written sentence saying what a rollback would mean for that agent today.
Preview, from the skill itself
“# Agent Config Versioning<br># Agent Config Versioning An agent is what its configuration says it is.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: 26ca48fb602c… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 41.4 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/the-broken-briefing/WALKTHROUGH.md, examples/the-policy-tightening/WALKTHROUGH.md, reference/rollback-procedure.md, reference/version-record.md, reference/version-storage-and-naming.md, reference/writing-the-diff.md, templates/rollback-record.template.md, templates/version-entry.template.md
Worked example, from the package
examples/the-broken-briefing/WALKTHROUGH.md
A solo operator runs a daily briefing agent that builds a morning briefing from his calendar, his task list, and his job notes. One afternoon change empties two sections of the next briefing. The version record names the change, the hour it went live, and the exact earlier text, and the rollback restores that text whole the same day.
Your recurring report comes out in the same shape every time, every figure traced to the record behind it, and a missing record named as a gap instead of quietly filled.
What changes
Honest fit
For you if: A report repeats on a schedule and its sections drift, its figures cannot be traced, or a missing record gets filled in from memory.
Skip it if: You need the underlying records created, a forecast made, or action taken on what the report shows; the builder works from records that exist and sends nothing on its own.
Your first run produces: one written spec for the recurring report you build most often, and one report built to it for the last completed period: every figure traced to its named record, every total re-added from the figures above it, and every missing record named as a gap.
Preview, from the skill itself
“# Scheduled Report Builder<br># Scheduled Report Builder Some reports are written once.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: 1e585d8790ad… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 36.4 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/the-monthly-cost-roundup/WALKTHROUGH.md, examples/the-weekly-operations-summary/WALKTHROUGH.md, reference/fixed-shape-and-sections.md, reference/gaps-and-missing-records.md, reference/review-and-approval.md, reference/tracing-figures-to-records.md, templates/report-spec.template.md, templates/scheduled-report.template.md
Worked example, from the package
examples/the-monthly-cost-roundup/WALKTHROUGH.md
An operator runs three standing agents and builds the same cost roundup in the first days of each month for the month just ended: spend by provider, spend by work type, and anything that cannot be placed. One provider's record has not arrived by the close day, one charge is grouped with no breakdown, and the roundup still closes with both facts in their sections and its totals labelled for what they are.
You can see every leg a secret travels between tools, agents, and people, each leg recorded with a plain risk call and a safer route, working from labels and locations only and never handling a real value.
What changes
Honest fit
For you if: A secret has to travel between tools, agents, or people, and nobody can say how it gets from its home to the place that uses it.
Skip it if: You need the inventory kept, an exposure ruling, or a value rotated; that is Secrets Hygiene work, and this check never holds, carries, or passes a real value.
Your first run produces: one filled transit path log for the secret that traveled most recently: one block per leg, a risk call on every leg, and totals counted from the rows, plus a draft safer route plan that changes nothing yet and a handover note to Secrets Hygiene for any value that traveled badly, in labels and dates only.
Preview, from the skill itself
“# Secrets in Transit Check<br># Secrets in Transit Check A secret at rest has a home.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: 97076d088b60… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 38.8 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/the-debug-handoff/WALKTHROUGH.md, examples/the-weekend-handover/WALKTHROUGH.md, reference/reading-transit-records.md, reference/risk-calls-on-a-path.md, reference/safer-routes.md, reference/where-secrets-get-copied.md, templates/safer-route-plan.template.md, templates/transit-path-log.template.md
Worked example, from the package
examples/the-debug-handoff/WALKTHROUGH.md
A small web project, one owner, one agent. A deploy fails on a Tuesday evening, and the two of them debug it together. The secret is the deploy service token. Its proper home, on the Secrets Hygiene inventory, is the hosting platform secret setting, but the setting sits empty: the token was reissued in a hurry during an outage two weeks earlier and parked in the owner's vault for now, and the deploy tool was set to take a typed value. This audit, run on Wednesday morning, follows the trip the token took on Tuesday night. It took about forty minutes.
While an incident is live, each outward update is drafted from the incident record in plain words, what happened, what is affected, and when the next update comes, and every draft is held for your approval before it goes out.
What changes
Honest fit
For you if: An incident is live and the people affected need what happened, what is affected, and when the next update comes, while the facts are still moving.
Skip it if: You need the incident fixed, the post incident review run, or one to one support replies handled; this skill drafts the outward words only, and sends nothing.
Your first run produces: a practice set from a past, closed incident: the incident record rebuilt from its own sources with gaps written in as unknown, a first notice drafted from only what was known at that early point and checked line by line against the record, and a closing note drafted from the end of the record, both held for the operator and logged as practice, not sent.
Preview, from the skill itself
“# Incident Comms Drafts<br># Incident Comms Drafts While an incident is live, the people affected by it do not need a full explanation.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: ea5430dec6b2… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 38.3 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/checkout-failure-live-updates.md, examples/sign-in-slowdown-live-updates.md, reference/corrections-and-closing.md, reference/plain-words-and-tone.md, reference/the-incident-record.md, reference/update-cadence-and-channels.md, templates/incident-record.template.md, templates/message-draft-set.template.md
Worked example, from the package
examples/checkout-failure-live-updates.md
A solo operator runs a small online shop that sells printed field guides. On a Tuesday morning checkout stops completing. Three messages carry the incident from first report to closing note, each drafted from one record and held for approval before it goes out.
Every beta report lands in one triaged record tied to the build it came from, duplicates merged and severity called from its effect, so every fix can be traced to the release that carried it and closed only when a reporter confirms it landed.
What changes
Honest fit
For you if: A beta build is out with testers and their reports arrive scattered, with no build on some and the same fault reported again and again.
Skip it if: You need testers recruited, fixes written and shipped, or a support desk run; the loop keeps and traces the reports, and the product calls stay yours.
Your first run produces: one record holding a whole beta round: every findable report logged with its build, each fault ranked by its effect, duplicates merged with their reporters kept, and any fix named only where the release record backs it, with nothing sent to testers.
Preview, from the skill itself
“# Beta Feedback Loop<br># Beta Feedback Loop Beta feedback fails in a quiet way.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: 784a948cb872… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 39.8 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/the-duplicate-wave/WALKTHROUGH.md, examples/the-export-crash/WALKTHROUGH.md, reference/logging-and-versioning.md, reference/merging-duplicates.md, reference/severity-by-effect.md, reference/tracing-fixes-to-releases.md, templates/feedback-record.template.md, templates/release-feedback-summary.template.md
Worked example, from the package
examples/the-duplicate-wave/WALKTHROUGH.md
A checklist app for small work crews is in beta with 8 testers, all of them crew leads using build 1.4.0 on real jobs. In the first 2 days, 11 reports arrive. Seven of them are one fault wearing seven sets of words. This walkthrough follows the merging, the one fix that shipped, and the block that honestly stays open.
A CSV becomes database rows under a column mapping you approved before anything loaded, bad rows set aside in a named rejects file instead of dropped, and a row-count reconciliation that proves the import complete.
What changes
Honest fit
For you if: A spreadsheet or export needs to become database rows, and past imports lost rows nobody could account for.
Skip it if: You need the file cleaned, the table reshaped, or rejected rows repaired for you; cleaning is separate work done first, and rejected rows go back to you whole.
Your first run produces: a validation pass that loads nothing: the smallest waiting file counted as received, its column mapping written and approved in the record, a named rejects file written, and the pass reconciled exactly, ready for the operator to read the mapping and the rejects and give the word to load.
Preview, from the skill itself
“# CSV to Database Import<br># CSV to Database Import A CSV file and a database table rarely line up on their own.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: 50e1891d876d… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 33.5 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/member-list-import/WALKTHROUGH.md, examples/supplier-parts-import/WALKTHROUGH.md, reference/loading-safely.md, reference/rejects-and-reconciliation.md, reference/validation-rules.md, reference/writing-the-mapping.md, templates/column-mapping.template.md, templates/import-record.template.md
Worked example, from the package
examples/member-list-import/WALKTHROUGH.md
A community workshop moves its membership list from a spreadsheet export into its members table: 486 rows in the file, 23 rejected with named reasons, 463 loaded, and both reconciliations closing exactly.
Tomorrow's job gets checked today: access confirmed, materials on hand counted against the job record, the customer confirmation drafted, weather glanced at, and anything short raised while there is still time to fix it.
What changes
Honest fit
For you if: One job runs tomorrow and its access, materials, and customer confirmation are still assumptions the morning would test.
Skip it if: You need a whole day of jobs planned, materials ordered, or a new time agreed with the customer; the checklist readies one job, and ordering and sending stay yours.
Your first run produces: one closed checklist for tomorrow's job: access read from the record, materials counted where they sit with needed, on hand, and short written per line, the weather glanced at with the time noted, every gap on one fix list with a name and a time, and one draft confirmation held for the operator to send.
Preview, from the skill itself
“# Pre-Job Site Checklist<br># Pre-Job Site Checklist The morning of a job is the wrong time to learn that a gate is locked, a part is short, or the customer expected a different day.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: c745e67aacfe… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 34.1 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/fence-and-gate-day-before/WALKTHROUGH.md, examples/roof-valley-repair-day-before/WALKTHROUGH.md, reference/access-and-entry.md, reference/customer-confirmation.md, reference/materials-count-against-the-record.md, reference/weather-and-day-fit.md, templates/confirmation-message.template.md, templates/site-checklist.template.md
Worked example, from the package
examples/fence-and-gate-day-before/WALKTHROUGH.md
A fencing operator with one job booked for Thursday: a side yard run of fence panels to replace, with a gate to hang at the end of the run. The check runs on Wednesday afternoon. There is one job on this sheet, and the crew is the operator and one helper, both named in the job record for this work.
Distance stops being priced from feel: your service area becomes named zones with a written trip line for each, stated the same way in every quote, and out-of-zone jobs handled as logged exceptions instead of quiet discounts.
What changes
Honest fit
For you if: Distance gets priced from feel, so near and far customers get trip figures nobody can explain later.
Skip it if: You need prices set, new ground added to the area, or quotes sent and negotiated; the figures and the area are yours, and exceptions stay yours to approve.
Your first run produces: one dated zone table built from the last run of finished jobs: three zones named with their edges written in one line each and a trip line stated per zone, the ground beyond named out of zone, and one real quote for this week written from the table, held for the operator's approval with nothing sent.
Preview, from the skill itself
“# Service Area Pricing Zones<br># Service Area Pricing Zones Distance is part of almost every local job, and it is the part most often priced from feel.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: fb1d3cc60a2c… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 33.9 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/cleaning-operator-named-zones/WALKTHROUGH.md, examples/hauling-operator-three-zones/WALKTHROUGH.md, reference/dividing-the-area-into-zones.md, reference/out-of-zone-exceptions.md, reference/setting-the-trip-line-per-zone.md, reference/stating-zones-in-quotes.md, templates/quote-and-exception-log.template.md, templates/zone-table.template.md
Worked example, from the package
examples/cleaning-operator-named-zones/WALKTHROUGH.md
A two person cleaning operator doing one off deep cleans and moving out cleans. Distance had been handled by adding a bit when a house sounded far, and nothing when it sounded near. The area is cut by a river: some addresses are close in miles and slow to reach, because every trip to them crosses one bridge. The operator zones by drive time instead of distance, from the job records, which hold the drive times.
An unhappy customer gets a ladder, not an argument: each rung from first contact to resolution has its record, its draft reply, and its limit, and no draft ever promises more than the job's written warranty.
What changes
Honest fit
For you if: A customer says work failed, came back, or was not what they expected, and the reply is becoming an argument instead of a path.
Skip it if: You need legal advice, replies sent for you, or refund and goodwill calls made; those decisions stay with you, and the ladder only keeps them written down.
Your first run produces: one complete complaint record for a current or recent complaint, logged in the customer's own words, with the job record pulled, the warranty check written as covered, not covered, or not clear with the dates and wording it rests on, and one reply drafted and held, promising nothing beyond the written warranty.
Preview, from the skill itself
“# Complaint Handling Ladder<br># Complaint Handling Ladder An unhappy customer gets a ladder, not an argument.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: 85594ab9676f… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 33.6 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/free-redo-asked-beyond-warranty/WALKTHROUGH.md, examples/leak-returned-inside-warranty/WALKTHROUGH.md, reference/closing-and-escalation.md, reference/reading-the-job-record.md, reference/reply-tone-and-limits.md, reference/the-written-warranty.md, templates/complaint-record.template.md, templates/reply-draft.template.md
Worked example, from the package
examples/free-redo-asked-beyond-warranty/WALKTHROUGH.md
A garden services operator who cleans and stains fences. Eleven months after a fence was cleaned and stained, the customer writes that the stain has gone patchy along the bottom rails and asks for the whole fence to be redone free, plus half the original price back for the trouble.
Finished jobs become a gallery you can actually show: before and after pairs pulled from each job's own photo record, captions that state only what the record supports, and the customer's permission noted on the record before any pair is used.
What changes
Honest fit
For you if: Finished jobs sit in a camera roll, unpaired and uncaptioned, with no note of which customers agreed to be shown.
Skip it if: You need photos taken or improved, permission asked for you, or pairs posted anywhere; the gallery is kept as a record, and showing anything stays yours.
Your first run produces: one complete gallery entry for a finished job with a good photo record: a matched before and after pair, a caption every word of which traces to the job record, and the permission note on the record with its date and scope, or a clear held mark and a drafted permission ask held for the operator to send.
Preview, from the skill itself
“# Before and After Gallery Keeping<br># Before and After Gallery Keeping A finished job is the clearest proof of work a local operator has, but only when the photos are paired honestly and the customer has agreed to them being shown.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: 58a0f7968179… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 27.5 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/deck-cleaning-pair-cleared-for-use/WALKTHROUGH.md, examples/roof-valley-repair-pair-held-back/WALKTHROUGH.md, reference/keeping-and-using-the-gallery.md, reference/pairing-photos-from-the-record.md, reference/permission-on-the-record.md, reference/writing-captions-from-the-record.md, templates/gallery-entry.template.md, templates/permission-ask.template.md
Worked example, from the package
examples/deck-cleaning-pair-cleared-for-use/WALKTHROUGH.md
A pressure washing and deck cleaning operator with a finished job: a wooden back deck cleaned and washed down, at a house in the operator's home town. The job record holds four photos and a short note of the work. This walkthrough takes that one job from photo record to a cleared gallery entry.
A multi-part series that agrees with itself: the arc and part list drawn from your real work, continuity notes kept for every part (what was said, what was promised), and each new part checked against them before you film.
What changes
Honest fit
For you if: A series is planned part by part from memory, and parts disagree on names, sizes, or promises made earlier.
Skip it if: You need the work invented, or filming, editing, and posting done; there is no series to plan without real work in progress, and the camera work stays yours.
Your first run produces: one series plan and one continuity ledger for a piece of real work that falls into parts: the arc in one sentence, the part list in the order the work happens, earlier parts filled into the ledger from the parts themselves, and the next part's outline checked line by line against it, held for the creator with nothing filmed or announced.
Preview, from the skill itself
“# Series and Sequels Planner<br># Series and Sequels Planner A series earns trust part by part.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: fc399db0d725… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 30.7 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/raised-bed-garden-series/WALKTHROUGH.md, examples/workbench-build-series/WALKTHROUGH.md, reference/arc-from-real-work.md, reference/checking-a-new-part.md, reference/continuity-notes.md, reference/sequels-and-returns.md, templates/continuity-ledger.template.md, templates/series-plan.template.md
Worked example, from the package
examples/raised-bed-garden-series/WALKTHROUGH.md
A grower films one raised bed from bare soil to first harvest. The season runs in five stages over four months, and the plants put in during the second part are the facts the fourth and fifth parts will be judged against. Memory alone will not hold which edge the beans went along, or how many pepper plants went in.
Sponsor outreach built on numbers you can stand behind: a one-page media sheet drawn from your own records, a pitch draft per prospect, and a sent and answered log so nobody gets pitched twice.
What changes
Honest fit
For you if: Sponsor outreach is being prepared and the figures in it come from memory, or a past round makes a repeat pitch likely.
Skip it if: You need rates set, deals promised, or sponsors spoken to; the skill drafts and keeps the log, and deciding, sending, and talking stay yours.
Your first run produces: one media sheet built from three records alone, the platform stats for the most recent full period, the current mail list count, and the count of items published, every figure labelled with its source and date, plus three prospect names checked against the log and one pitch draft held for the operator, with nothing sent.
Preview, from the skill itself
“# Sponsorship Outreach Drafts<br># Sponsorship Outreach Drafts A sponsor pitch is only as good as the numbers in it, and the numbers are only good if they come from the creator's own records.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: 134671316d45… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 28.9 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/cooking-channel-return-round/WALKTHROUGH.md, examples/podcast-first-sponsor-round/WALKTHROUGH.md, reference/building-the-prospect-list.md, reference/drafting-the-pitch.md, reference/media-sheet-from-records.md, reference/the-sent-and-answered-log.md, templates/media-sheet.template.md, templates/pitch-draft.template.md
Worked example, from the package
examples/cooking-channel-return-round/WALKTHROUGH.md
A creator who posts short videos on weeknight cooking for small households. A first outreach round ran earlier in the year. This is the second round, and its whole point is discipline: fresh figures, and not one prospect from the first round pitched again as if that round never happened.
A new workflow earns its first live run: it rehearses against copies of your real records and hands you a step-by-step log with every outward or state-changing step flagged, so you approve exactly what it would do before it touches anything real.
What changes
Honest fit
For you if: A new workflow is about to run live for the first time, and its first real mistake would be costly or hard to undo.
Skip it if: You need the workflow itself judged correct; a rehearsal shows what it would do on the records copied, and the wording, pricing, and judgement inside its steps stay yours to settle.
Your first run produces: one rehearsal of the next new workflow you were about to run live: a one-screen brief, dated copies of three records it would really meet (one plain case, one with a missing or odd field, one near a limit), a step-by-step log with every outward and state changing step flagged, and your decision recorded on the log, approved, approved with corrections, or not approved.
Preview, from the skill itself
“# Agent Dry Run Rehearsal<br># Agent Dry Run Rehearsal A new workflow looks safest on the day it is written.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: dc7f40f1c6c7… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 36.5 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/the-invoice-chase-rehearsal/WALKTHROUGH.md, examples/the-job-file-tidy-rehearsal/WALKTHROUGH.md, reference/copying-records-safely.md, reference/flagging-steps.md, reference/reading-the-rehearsal-log.md, reference/when-a-rehearsal-fails.md, templates/rehearsal-brief.template.md, templates/rehearsal-log.template.md
Worked example, from the package
examples/the-invoice-chase-rehearsal/WALKTHROUGH.md
A small studio is about to hand its overdue invoice chase to a new workflow for the first time. The workflow reads the ledger, picks out overdue invoices, drafts a chase note for each one, sends it, and marks the invoice as chased with a follow up date. Every one of those steps after the reading is outward or state changing, so the first run is a rehearsal on copies, and the live chase waits for the log.
Estimates stop being guesses: every task's estimate is written before it starts and its actual time beside it when it ends, so the next estimate for the same work comes from recorded actuals and chronic misses get named.
What changes
Honest fit
For you if: The same kinds of work keep taking longer, or shorter, than the time you plan for them, and nothing is written down to prove it.
Skip it if: You need tasks made faster, people scored, or clients billed from the minutes; the record holds time honestly, and what changes about the work stays your call.
Your first run produces: one time record for a kind of work you do at least twice a week, holding three tasks of that kind with the estimate written before each one starts and the actual beside it when it ends, and a fourth estimate set from those recorded actuals instead of habit.
Preview, from the skill itself
“# Task Time Estimates<br># Task Time Estimates Most work starts with a number nobody wrote down.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: 2744db7b1013… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 34.7 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/the-inbox-that-always-ran-long/WALKTHROUGH.md, examples/the-release-notes-that-came-in-early/WALKTHROUGH.md, reference/estimate-records.md, reference/handling-splits-and-interruptions.md, reference/making-the-next-estimate.md, reference/naming-chronic-misses.md, templates/estimate-review.template.md, templates/time-record.template.md
Worked example, from the package
examples/the-inbox-that-always-ran-long/WALKTHROUGH.md
A solo operator starts each morning with inbox triage and always plans twenty minutes for it. The work always takes longer, the morning always starts late, and the estimate never changes, because nothing was ever written down. Six tasks in one record end that pattern.
Agent handoffs stop breaking two steps later: every handoff format is written down as a contract (fields, order, what must never appear in it), and every output is checked against that contract before anything downstream uses it.
What changes
Honest fit
For you if: Agents hand work to each other, and downstream steps keep breaking on a missing field, a field in the wrong place, or material that should never travel inside a handoff.
Skip it if: You need the content inside a well shaped output checked for truth; this skill checks shape and completeness only, and a passing output can still carry a wrong figure.
Your first run produces: one contract for the handoff that has caused the most rework, written on one screen with its fields in order and its forbidden list, the last real output in that format checked against it by hand with every fault counted and named, and the contract agreed, dated, and filed where both sides of the handoff can read it.
Preview, from the skill itself
“# Output Format Contracts<br># Output Format Contracts Agents hand work to each other and assume the shape.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: e5b345ec340e… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 33.5 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/the-invoice-handoff/WALKTHROUGH.md, examples/the-job-status-handoff/WALKTHROUGH.md, reference/changing-a-contract.md, reference/checking-an-output.md, reference/forbidden-content.md, reference/writing-a-contract.md, templates/conformance-report.template.md, templates/format-contract.template.md
Worked example, from the package
examples/the-invoice-handoff/WALKTHROUGH.md
A drafting agent prepares invoice figures and hands them to a sending agent, which builds the client invoice from them. The first handoff under the new contract is held with three faults. The fixed handoff passes a full recheck. All figures in this walkthrough are internal example figures, used only to show the check working.
No renewal lands as a surprise charge: every subscription and tool sits in one record with its renewal date, its price, and last time's price beside it, and the keep-or-drop call is recorded before the money goes.
What changes
Honest fit
For you if: Subscriptions and tools renew across the year, and nobody can say from one record what the studio pays for, when, or at what price.
Skip it if: You need bills paid, plans changed, or prices negotiated; the skill keeps the record and prepares the call, and approving and acting stay yours.
Your first run produces: one renewal record holding every subscription, tool, and service the studio pays for, built from billing sources rather than memory, each item with its renewal date and its price this time and last time or its gap named, and a keep-or-drop call with a reason beside every item due in the next sixty days, with that window's total written at the top.
Preview, from the skill itself
“# Vendor and Tool Renewals<br># Vendor and Tool Renewals A studio pays for its tools one quiet charge at a time.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: de5189866fd4… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 40.9 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/the-look-ahead-review.md, examples/the-price-rise-call.md, reference/review-rhythm-and-alerts.md, reference/sources-prices-and-terms.md, reference/the-keep-or-drop-call.md, reference/the-renewal-record.md, templates/renewal-record.template.md, templates/renewal-review.template.md
Worked example, from the package
examples/the-look-ahead-review.md
A small studio keeps one renewal record with 7 active entries. The monthly review runs on March 3. Its window is March 3 to May 2, the next sixty days. Four items fall due in the window. The other 3 renew after the window: a hosting plan on June 16, a backup service on June 20, and an accounting tool on September 5. They are listed by date only.
Your published docs stop quietly going stale: on a schedule, every claim in a guide or README is traced to the shipped build, drift is logged with the exact line it lives on, and each drifted line is fixed or cut.
What changes
Honest fit
For you if: Published guides or READMEs were true when written, and the product has shipped changes since.
Skip it if: You need a guide written in the first place, or the product itself fixed; the check starts from a published doc and hands over a checked packet, and publishing the correction stays yours.
Your first run produces: one claim inventory for the published guide or README your readers open most, with the first ten claims traced against the current build and marked holds or drifted, every drifted claim logged with its line, the doc's words, and the build's words, and one fix drafted in the doc's own style and held for the operator.
Preview, from the skill itself
“# Doc Drift Checks<br># Doc Drift Checks A published guide does not go wrong all at once.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: a35b95fb113f… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 44.3 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/guide-drift-after-release/WALKTHROUGH.md, examples/readme-drift-after-rename/WALKTHROUGH.md, reference/claim-inventory-and-tracing.md, reference/drift-kinds-and-severity.md, reference/fix-or-cut-decisions.md, reference/scheduling-and-baselines.md, templates/claim-inventory.template.md, templates/drift-log.template.md
Worked example, from the package
examples/guide-drift-after-release/WALKTHROUGH.md
A small invoicing tool, LedgerPad, ships a new build. Its published Getting Started guide was true at publication. A release triggered check traces all 36 of its claims against the new build and finds 6 drifted, spread across the guide, in a release whose notes mentioned only one of them.
Forms stop carrying figures from memory: every field is mapped to a named record before filling, values come from that record only, a check pass reads the filled copy back against the records, and the submitted copy is kept with its date.
What changes
Honest fit
For you if: Forms get filled from memory, and the figures on them cannot be traced to a record afterwards.
Skip it if: You need answers chosen, a signature applied, or advice on what a form means; the skill marks those fields for you and fills only from the records in the mapping.
Your first run produces: one field mapping for the short form that is waiting right now, one row per field naming the record and the place in it, a working copy filled from that mapping with values copied exactly, and a check pass read back against the records, with the filled copy, the check result, and any gaps shown to the operator before anything is submitted.
Preview, from the skill itself
“# PDF Form Filling<br># PDF Form Filling A blank form asks questions the operator's records can usually answer.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: 3be693493d71… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 40.2 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/event-permit-application.md, examples/vendor-registration-form.md, reference/field-mapping.md, reference/keeping-the-submitted-copy.md, reference/reading-the-records.md, reference/the-check-pass.md, templates/field-mapping.template.md, templates/submission-record.template.md
Worked example, from the package
examples/event-permit-application.md
A makers group applies to the city for a one day market in a public park. The permit form is two pages and nine fields, the facts live in three short records the group keeps, and the check pass earns its place by catching the event date typed one day early, a mistake the site letter confirms as a mistake.
A missed delivery stops being an argument: the day is planned from the job records with load order and stop order written down, a customer message drafted per stop, and a proof-of-drop note at every stop, so any miss traces to the plan or the road.
What changes
Honest fit
For you if: Delivery days are planned in the driver's head, and a missed drop turns into an argument from memory.
Skip it if: You need the driving, loading, or dropping done, or messages sent; the skill plans the day and writes the proof record, and the road work stays yours.
Your first run produces: one run plan for the next delivery day on the books, built from the job records only: each stop with what it is owed, its address, and any drop note, gaps flagged in plain words, the stop order set and the load order set in reverse, and one draft message per stop held for the operator, with nothing sent and nothing loaded.
Preview, from the skill itself
“# Delivery and Dropoff Runs<br># Delivery and Dropoff Runs A delivery day goes wrong in quiet ways.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: b88031f6da78… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 29.5 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/five-stop-round-missed-delivery/WALKTHROUGH.md, examples/six-stop-firewood-round/WALKTHROUGH.md, reference/customer-messages-per-stop.md, reference/load-order-and-stop-order.md, reference/proof-of-drop-and-tracing.md, reference/reading-the-job-records.md, templates/proof-of-drop-log.template.md, templates/run-plan.template.md
Worked example, from the package
examples/five-stop-round-missed-delivery/WALKTHROUGH.md
An operator delivering bagged mulch and bagged compost to garden customers in one loop. Five stops are booked for the day, all bagged, all able to ride in one pickup load with the piles kept apart by count. One stop is not delivered, and the day is closed by tracing that miss to the road, not to the plan.
Returns and damage questions get a record instead of an argument: every rented or borrowed item logged with what it is, whose it is, the deposit and due-back date, and its condition noted out and back with photos from the job record.
What changes
Honest fit
For you if: Rented or borrowed gear is out on jobs, and its deposits, due-back dates, and condition live on slips and in memory.
Skip it if: You need gear booked, deposits paid, or damage decided; the log lays out the notes, photos, and receipts, and settling the question stays with you and the owner.
Your first run produces: one gear log holding every item the business has out right now that it does not own, rented, hired, or borrowed: what it is, whose it is, the deposit and due-back date from the paperwork or agreement or an honest none or unknown, and its condition as it stands today, marked as a mid hire note where the gear came in before the log started.
Preview, from the skill itself
“# Rental and Borrowed Gear Log<br># Rental and Borrowed Gear Log Gear that is not yours still has to go back.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: 6ae5651d82ca… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 36.4 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/borrowed-scaffold-tower-house-job.md, examples/rented-mini-excavator-drainage-job.md, reference/check-out-and-condition-notes.md, reference/deposits-due-dates-and-fees.md, reference/gear-log-fields.md, reference/return-check-and-damage-questions.md, templates/check-out-return-entry.template.md, templates/gear-log.template.md
Worked example, from the package
examples/borrowed-scaffold-tower-house-job.md
A painter and exterior repair operator needs a scaffold tower and a long ladder for a two storey house job. A neighbouring roofer the operator has worked beside for years lends both for the week. There is no rental firm and no hire paperwork. The gear is not owned by the operator, so it goes in this log on the same terms as rented gear, with the loan terms as agreed written in place of paperwork.
The week's money in is proven, not assumed: cash, Cash App, and card receipts matched line by line against the job records, every gap named against the job it belongs to, and nothing written off without your recorded decision.
What changes
Honest fit
For you if: A week's jobs are done, and whether every one of them was paid in full is a feeling rather than a match.
Skip it if: You need bookkeeping or tax advice, money moved, or a week judged profitable; the sheet matches money in against the job records and presents nothing as profit.
Your first run produces: one reconciliation sheet for the last finished week: the expected list built from the job records alone and totalled before any payment record is opened, then cash, Cash App, and card lines matched against it, ending with the expected total, the matched total, and a gap list where every gap names its job, held for the operator to read.
Preview, from the skill itself
“# Receipts Reconciliation<br># Receipts Reconciliation At the end of a busy week, an operator can know the work was done and still not know whether they were paid for all of it.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: 90dd40283505… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 33.7 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/five-jobs-one-short-payment.md, examples/split-payments-fees-and-one-unknown.md, reference/gaps-named-never-written-off.md, reference/matching-line-by-line.md, reference/reading-payment-records.md, reference/the-job-record-baseline.md, templates/gap-note.template.md, templates/weekly-reconciliation-sheet.template.md
Worked example, from the package
examples/five-jobs-one-short-payment.md
A local operator closes a week with five jobs on the records: gutter work, a fence repair, a yard cleanup, a limb removal, and a driveway wash. Payment came in by cash, Cash App, and card across the week. This walkthrough is the weekly reconciliation for that week.
Business-to-business referrals stop depending on memory: one working list of who sends what kind of work, what has been sent each way, the thank-you kept current, and one nudge drafted when a good partner has gone quiet.
What changes
Honest fit
For you if: Work already moves between your business and others, and who sent what, and who is owed a thank-you, lives in your head and old threads.
Skip it if: You need new partners found, customers asked for referrals, or thank-yous and nudges sent; the list records the trade between businesses, and sending stays yours.
Your first run produces: one partner list built from the last season of job records, one entry per business the operator really traded a referral with, each entry holding the fit in both directions and a tally that traces to logged lines or is marked as starting today, the quiet window in the list header, and one plainly owed thank-you drafted and held for the operator, with nothing sent.
Preview, from the skill itself
“# Referral Partner List<br># Referral Partner List Most local operators already trade work with other local businesses.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: e8150d6938e8… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 38.8 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/cleaner-agent-circle-and-a-quiet-partner/WALKTHROUGH.md, examples/roofer-plumber-two-way-trade/WALKTHROUGH.md, reference/thank-yous-that-stay-current.md, reference/the-partner-list.md, reference/what-counts-as-a-referral.md, reference/when-a-partner-goes-quiet.md, templates/nudge-and-thank-you.template.md, templates/partner-entry.template.md
Worked example, from the package
examples/cleaner-agent-circle-and-a-quiet-partner/WALKTHROUGH.md
A cleaning operator who does move out cleans, end of tenancy cleans, and pre sale cleans. The trade runs in a small circle: a letting agent, a removals firm, a window cleaning firm, and a plumbing and heating firm whose landlord work used to swap steadily with hers. A review covers twelve weeks, early February to late April.
A live session starts planned instead of improvised: the run order written from your real work, the setup checked before you go live, and the list of clips and follow-ups made before the session ends, not from memory after.
What changes
Honest fit
For you if: Live sessions start improvised, and the clips and follow-ups get rebuilt from memory after they end.
Skip it if: You need the session hosted, recorded, or edited, or content invented to fill it; the plan is built from real work only, and a shortfall means a shorter session, not a padded one.
Your first run produces: one complete session plan for the next live session already on the calendar: a brief taken from the creator's real work, a run order whose block times add up to the planned length, a clip and follow-up list drafted from that run order, and a setup check with every line marked, all held for the creator with nothing announced.
Preview, from the skill itself
“# Live Session Prep<br># Live Session Prep A live session goes better when it is planned while there is still time to fix things.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: e1ff5d1a1a51… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 33.1 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/sketching-tutorial-live/WALKTHROUGH.md, examples/workshop-demo-live/WALKTHROUGH.md, reference/planning-clips-and-follow-ups.md, reference/the-setup-check.md, reference/working-from-real-work.md, reference/writing-the-run-order.md, templates/session-plan.template.md, templates/setup-check.template.md
Worked example, from the package
examples/sketching-tutorial-live/WALKTHROUGH.md
An artist runs a live tutorial drawn from their own sketchbook work: one half finished drawing of a kitchen chair, made by the artist in the week before the session, is shaded and finished live while beginners follow the steps. The drawing is the real work. A fresh page sits beside it for the first sketch step, so the half finished piece is not touched until its block.
Your next filming list comes from your audience's own words: real questions and requests from comments logged verbatim, grouped by theme, and every filming item naming the comments that asked for it.
What changes
Honest fit
For you if: Your audience's questions and requests sit in comment piles, and the next filming list gets built from memory instead.
Skip it if: You need comments replied to, videos filmed or posted, or topics brought in from trends; the log holds only the comments you supply, and logging an idea is not a promise to any viewer.
Your first run produces: one comment log for a recent video with a modest comment pile: the batch counted before logging, each comment holding a question or request logged in the commenter's own words, the rest set aside with reasons, the log grouped by theme, and a short draft filming list held for the creator where every item names the comments behind it.
Preview, from the skill itself
“# Comment to Content Ideas<br># Comment to Content Ideas The next video is often already written in the comments.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: ffbf204e0cc0… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 33.4 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/beginner-baking-follow-ups/WALKTHROUGH.md, examples/workshop-tool-questions/WALKTHROUGH.md, reference/building-the-filming-list.md, reference/grouping-by-theme.md, reference/keeping-the-idea-log-current.md, reference/logging-comments-verbatim.md, templates/comment-idea-log.template.md, templates/filming-list.template.md
Worked example, from the package
examples/beginner-baking-follow-ups/WALKTHROUGH.md
A home baking creator who posts plain, beginner level bakes. One video, a basic white loaf mixed and baked in a single session, drew comments from people who tried the loaf and hit the same points. The creator wants the follow up filming list to come from those tries, not from a guess about what beginners need.
A model or tool change stops shipping silent surprises: one fixed set of standard tasks runs on the old setup and the new one before anything goes live, every difference is named in plain words, and the rollback note is already written if the swap has to come back out.
What changes
Honest fit
For you if: You run an agent whose model or a tool it relies on is about to change, and a silent behaviour change would be found late, in real work.
Skip it if: You need the new model or tool chosen for you, differences fixed, or a setup proven correct in general; the task set shows behaviour on its tasks only, and choosing and fixing stay with you.
Your first run produces: one fixed task set for the agent most likely to change next: 8 to 12 real tasks with their inputs written in full, the current setup named in full as the standing baseline, one kept baseline run of the set on that setup, and a one sentence rollback note that can be carried out.
Preview, from the skill itself
“# Model and Tool Swap Checks<br># Model and Tool Swap Checks An agent is only partly its instructions.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: 68814a48c0cc… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 41.6 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/the-model-upgrade/WALKTHROUGH.md, examples/the-search-tool-change/WALKTHROUGH.md, reference/reading-the-differences.md, reference/rollback-note-and-restore.md, reference/the-standard-task-set.md, reference/the-swap-record.md, templates/rollback-note.template.md, templates/swap-check.template.md
Worked example, from the package
examples/the-model-upgrade/WALKTHROUGH.md
A solo operator runs an inbox triage agent that labels each incoming message, holds or passes each one on, and drafts a short summary line for the held ones. A newer model is offered for the agent. The fixed task set is run on both models before anything goes live. The comparison finds three substance differences, one of them a fault, and the swap is held the same day, with the old model still live throughout.
Standing instructions stop quietly going stale: on a schedule you set, every rule is checked against the record of what actually happened and marked holding, stale, in conflict, or without evidence, with any change drafted for your approval and nothing edited silently.
What changes
Honest fit
For you if: Your agent's standing instructions were written for an older setup and have not been read against the run record in a while.
Skip it if: You need version history for instruction changes, new instructions written from nothing, or the agent graded as a whole; this review marks and drafts, and you approve before any wording changes.
Your first run produces: one review sheet for the agent whose instructions you have read least recently: the full instruction text copied exactly as it stands, each rule marked holds, stale, in conflict, or no evidence against the last two to four weeks of record, change drafts for the findings that call for them, and the next review date set before the sheet closes.
Preview, from the skill itself
“# Standing Instructions Review<br># Standing Instructions Review Standing instructions are written at one moment, for the setup as it stood at that moment.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: caa3e32f92e9… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 44.5 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/the-conflicting-reply-rules/WALKTHROUGH.md, examples/the-stale-hold-rule/WALKTHROUGH.md, reference/drafting-the-change.md, reference/finding-staleness-and-conflict.md, reference/reading-the-run-record.md, reference/review-record.md, templates/change-draft.template.md, templates/review-sheet.template.md
Worked example, from the package
examples/the-conflicting-reply-rules/WALKTHROUGH.md
A small shop runs a customer message agent that answers routine questions on its own and holds the rest for the operator. Two standing rules each read plainly on their own. Together they give two answers to the same message, and the run record shows the agent splitting between them for three weeks. The review names the pair, counts the split, and carries the operator choice back into the wording itself.
Records stop being kept forever by default: every record type gets a written keep period and review date, and the deletion pass actually runs on schedule, under its own approval, with every pass logged, so what is kept and when it last went can always be answered.
What changes
Honest fit
For you if: Your records are kept forever by default and nobody can say how long each type should stay or when it last went.
Skip it if: You need day to day data handling rules, consent tracking, or the keep periods chosen for you; the periods are your decision, and no pass runs without your approval.
Your first run produces: one starting schedule for the store where old records pile up most: every record type listed in plain words, each counted from the store itself today, a keep period per type or an honest not yet set, a review date per type, and the schedule read back against the store; no pass runs on a first run.
Preview, from the skill itself
“# Data Retention Schedule<br># Data Retention Schedule Records do not age out on their own.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: 1a335c91f371… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 33.4 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/first-schedule-cleanup/WALKTHROUGH.md, examples/quarterly-deletion-pass/WALKTHROUGH.md, reference/deletion-pass-procedure.md, reference/record-types-and-periods.md, reference/retention-schedule-record.md, reference/review-and-logging.md, templates/deletion-pass-log.template.md, templates/retention-schedule.template.md
Worked example, from the package
examples/first-schedule-cleanup/WALKTHROUGH.md
A solo operator keeps the records of a small briefing service in one store: task logs, draft briefings, exported reports, and intake notes. Nothing has a stated end. The first schedule lists the four types, sets a period and a review date for each, and the first deletion pass takes only what the first check found past its period, on a list the operator approved, and the pass is logged.
A published mistake stops being a quiet edit nobody can trace: the original is fixed in place, then one plain-words log entry per correction is published and kept, saying what was wrong, what it says now, and where it was fixed.
What changes
Honest fit
For you if: Your studio publishes in public and wants a settled, provable way to handle the mistakes that will sooner or later stand in it.
Skip it if: You need the right fact decided for you, apologies or client statements written, or private complaints tracked; the log covers errors that stood in public, and every entry waits for your approval.
Your first run produces: one public corrections log set up before the next error needs it: one public place a reader can reach in one step, a dated heading in plain words, a link from the places readers already land, the entry shape ready, and one practice entry only if a past fix can be proved from records.
Preview, from the skill itself
“# Public Corrections Log<br># Public Corrections Log Every studio publishes something wrong sooner or later.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: 435d9ec37ec1… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 37.1 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/the-wrong-date/WALKTHROUGH.md, examples/the-wrong-price/WALKTHROUGH.md, reference/the-log-itself.md, reference/verifying-the-fix.md, reference/where-the-log-lives.md, reference/writing-the-correction.md, templates/correction-entry.template.md, templates/published-correction-note.template.md
Worked example, from the package
examples/the-wrong-date/WALKTHROUGH.md
A small studio announces a live demo session in a post, on its schedule page, and in a short notice. All three carry a date typed from memory, one week early. The booking record holds the right date. All three places are fixed, and one log entry covers the correction, with a short note placed where the format allows one.
Retiring a product or a version stops being a mess: the decision and stop date are recorded, the sunset notice is drafted and held for your approval, what keeps working and until when is stated plainly, the records are exported and checked before anything stops, and the stop date is honored exactly.
What changes
Honest fit
For you if: You have a product or a version in mind to retire and do not want its ending reconstructed from memory later.
Skip it if: You need the retirement decided for you, a replacement named or priced, or users moved to one; this skill runs a decision honestly once you have made it.
Your first run produces: one sunset record for the product or version you have in mind to retire: the product named as its users know it, the decision written in your own words with its reason, every part of the product listed with what it does, and one sentence naming where its records would be exported to; nothing is stopped and nothing outward facing is drafted.
Preview, from the skill itself
“# Archive and Sunset Products<br># Archive and Sunset Products Every product ends, and every version is one day replaced.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: 0306a635698e… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 42.5 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/the-clean-retirement/WALKTHROUGH.md, examples/the-version-sunset/WALKTHROUGH.md, reference/export-and-retention-checklist.md, reference/sunset-record.md, reference/the-stop-date-and-final-shutdown.md, reference/writing-the-sunset-notice.md, templates/sunset-notice.template.md, templates/sunset-record.template.md
Worked example, from the package
examples/the-clean-retirement/WALKTHROUGH.md
A solo operator retires a small studio tool in full. The decision, the parts, and the dates are recorded first. The first export check falls short and is fixed while the tool still runs. The notice is approved before it goes out, the stop date holds, and the close is written into the record the day the last part stops.
A spreadsheet becomes a report a reader can act on: every figure in the prose is traced to the sheet's own cells, every total is recomputed from those cells instead of copied, and anything the sheet does not support is left out or named as missing.
What changes
Honest fit
For you if: You hold figures in a sheet and need a written report a reader can act on, with every figure answerable to its cells.
Skip it if: You need the sheet tidied or repaired, new data gathered, or the business behind the numbers judged; the report uses only what the sheet supports and names the rest as missing.
Your first run produces: one short report from a small sheet you already trust: the sheet named with its period and source, three figures traced to their cells with the total and the difference recomputed, a draft that uses only those figures, one gap the sheet does not hold named plainly, and the draft and trace held for the operator's check.
Preview, from the skill itself
“# Spreadsheet to Report<br># Spreadsheet to Report A spreadsheet holds numbers.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: f82a3a3e5263… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 33.7 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/project-costs-summary/WALKTHROUGH.md, examples/quarterly-sales-summary/WALKTHROUGH.md, reference/checking-the-report.md, reference/reading-the-sheet.md, reference/tracing-and-recomputing.md, reference/writing-the-report.md, templates/figure-trace.template.md, templates/report-outline.template.md
Worked example, from the package
examples/project-costs-summary/WALKTHROUGH.md
A builder finishes a small job and needs a written cost summary for the customer: what was budgeted, what was spent, and where the two differ. The figures sit in one sheet with a budget column and an actual column. The report is built by tracing each line and recomputing both totals and every difference, and it leaves out the one point the sheet cannot carry.
Tomorrow's tools and materials get staged tonight in job order, with every gap flagged and filled while the shops are still open, so the morning starts with driving to the first job instead of fetching.
What changes
Honest fit
For you if: Your mornings start with fetching because tomorrow's tools and materials were never checked against the truck the night before.
Skip it if: You need the job site itself checked, materials or quantities chosen, or stock tracked across weeks; staging covers what the crew carries, and buying stays with you.
Your first run produces: one staging sheet for a real next day, built the evening before at the truck and the bench: one block per job in run order with the tools and materials each job needs, every line marked on hand, bought, found, or carried open with a stated plan, gaps on a buy list while the shops are still open, and the load put on in job order.
Preview, from the skill itself
“# Tool and Material Staging<br># Tool and Material Staging Most lost mornings are lost the night before.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: 46db52eb08b8… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 34.7 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/roof-and-bathroom-staging/WALKTHROUGH.md, examples/three-job-tuesday-staging/WALKTHROUGH.md, reference/bench-check-and-close-out.md, reference/building-the-staging-list.md, reference/flagging-and-filling-gaps.md, reference/loading-in-job-order.md, templates/gap-and-buy-list.template.md, templates/staging-sheet.template.md
Worked example, from the package
examples/roof-and-bathroom-staging/WALKTHROUGH.md
A Wednesday with two jobs, staged on the Tuesday evening before. The job records list a roof valley repair first, in the morning, and a bathroom reseal second, in the afternoon. This walkthrough shows a staging that ends honestly short on one line, with the gap carried open on a stated plan, so the day still starts on time and nobody is surprised at the first address.
Every job gets closed out the same day in one log, with what was done, what is left, the photos filed against the job, and the customer's next step noted, so tomorrow starts from the log instead of from memory.
What changes
Honest fit
For you if: Your days touch more than one job and end fast, so what was done and what is left lives in your head by morning.
Skip it if: You need the final handover for a finished job, prices or pay worked out, or customer notes sent; the daily log is the closeout record, and the handover is a separate skill.
Your first run produces: one daily log for a real working day: every job touched that day listed from the job records, an entry per job with what was done and what is left, the day's photos filed against their jobs and counted, a next step on every entry, and the header hours and photo counts agreeing with the entries, read by the operator before it is filed.
Preview, from the skill itself
“# End of Day Job Log<br># End of Day Job Log A working day touches more than one job, and it ends fast.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: c8fdb2f3db97… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 32.2 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/deck-rebuild-day-two/WALKTHROUGH.md, examples/three-job-cleanup-day/WALKTHROUGH.md, reference/building-the-daily-log.md, reference/filing-the-days-photos.md, reference/next-steps-and-tomorrow-start.md, reference/what-was-done-and-what-is-left.md, templates/daily-job-log.template.md, templates/job-closeout-entry.template.md
Worked example, from the package
examples/deck-rebuild-day-two/WALKTHROUGH.md
A deck rebuild planned over 3 days. This walkthrough is the daily log for day 2 only. One job was touched that day, and it is still in progress at the end of day 2, so the log lists 1 job: 0 finished that day and 1 still in progress, and the 0 plus the 1 adds to the 1 job logged. Hours on the job on day 2 are 6.5 hours. Photos filed on day 2 are 7 photos. This is a daily closeout for a job in progress. It is not a handover, and no handover step is taken in it.
Every gift card, rain check, and store credit sits in one record, issued to whom, with its balance worked out redemption by redemption, so a credit is honored from the record and never argued from memory.
What changes
Honest fit
For you if: You issue gift cards, rain checks, or store credit and balances get argued from memory at the counter.
Skip it if: You need prices, credit amounts, or expiry terms set for you, payment taken, or legal advice on credit rules; the record holds the terms you set, exactly as set.
Your first run produces: one credit record started with the next credit you issue: the holder, the kind of credit, the amount issued, the date, and the expiry term you set, written down at the counter while both sides hold the facts, with a redemption log opened and a balance that follows from the entry itself.
Preview, from the skill itself
“# Gift Card and Credit Tracking<br># Gift Card and Credit Tracking A credit owed to a customer is easy to lose.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: 9a2b0eec87f0… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 29.1 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/gift-card-full-redemption/WALKTHROUGH.md, examples/store-credit-and-rain-check/WALKTHROUGH.md, reference/expiry-and-disputes.md, reference/issuing-a-credit.md, reference/redeeming-and-balances.md, reference/the-credit-record.md, templates/credit-record.template.md, templates/redemption-log.template.md
Worked example, from the package
examples/gift-card-full-redemption/WALKTHROUGH.md
A gift card for a cleaning service, bought as a gift and spent over three visits. The record holds one credit only, reference GC-014, issued for $120.00. Every figure in this walkthrough comes from that one entry, and the balances follow line by line. This walkthrough shows the full life of the credit, from issue to a closing balance of zero.
Public reviews get answered from the record instead of in a hurry: each review read in full, every point checked against the job's own facts, the reply drafted without inventing anything, and every draft held for your approval before it is posted.
What changes
Honest fit
For you if: Public reviews of your work need answers, and replies typed in a hurry from memory keep adding facts nobody checked.
Skip it if: You need reviews asked for, replies posted for you, or a dispute settled; every draft here is checked against the job record and waits for your approval before it posts.
Your first run produces: one review response draft for the next review that arrives: the full review text opened with its rating and date as shown, each point in it checked against the job record or marked as a gap, a draft using only record facts, and the draft held with its check log for your decision to post, edit, or leave it unposted.
Preview, from the skill itself
“# Review Response Drafts<br># Review Response Drafts A public review sits where the next customer will read it, and a reply written in a hurry can do more harm than the review itself.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: e7b9b270f8a4… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 31.7 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/hard-fence-repair-review/WALKTHROUGH.md, examples/positive-yard-cleanup-review/WALKTHROUGH.md, reference/checking-against-the-job-record.md, reference/drafting-from-the-record.md, reference/hard-reviews-and-approval.md, reference/reading-the-review-in-full.md, templates/review-check-log.template.md, templates/review-response-draft.template.md
Worked example, from the package
examples/hard-fence-repair-review/WALKTHROUGH.md
This is a fictional worked example. The review text, the names, and the figures are made up for this example, and they tie to each other only inside it.
Title and thumbnail picks stop being last-minute guesses: variants are written down before publishing, the pick is reviewed after a set window against the video's own numbers, and the outcome is logged so the next choice starts from what the record shows.
What changes
Honest fit
For you if: Your title and thumbnail picks happen in the last minutes before publishing and the reasons are gone by the next video.
Skip it if: You need growth promised, figures from other channels used as targets, or videos published for you; the log works only from your own channel's record.
Your first run produces: one variant sheet for the next video you have not published yet: a test log started from your channel's own record, the title and thumbnail variants for the new video written down before publishing, the pick and its reason noted from the log, the review window set on the sheet, and the pick approved by you before anything goes out.
Preview, from the skill itself
“# Thumbnail and Title Tests<br># Thumbnail and Title Tests A title and a thumbnail get picked in a hurry, often in the last ten minutes before a video goes out.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: 7cb3a7c0a94b… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 38.9 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/three-dinners-from-one-chicken/WALKTHROUGH.md, examples/workbench-from-reclaimed-timber/WALKTHROUGH.md, reference/reading-the-review-window.md, reference/the-test-log.md, reference/what-the-record-can-and-cannot-show.md, reference/writing-variants-before-publishing.md, templates/review-log.template.md, templates/variant-sheet.template.md
Worked example, from the package
examples/three-dinners-from-one-chicken/WALKTHROUGH.md
This is a fictional channel and a fictional video, shown to walk the workflow through in full, including a result that comes in below the channel's past entry. Mara Cooks Small is a small cooking channel. All figures below are that channel's own internal record for the purpose of this example, and each rate is worked from the impressions and clicks on its own line. No figure here comes from another channel and none is presented as typical for any platform.
Guests run as a pipeline instead of a scramble: prospects drawn from your own network and past episodes, tracked by stage, outreach drafted and held for your approval, a prep sheet built from the guest's public work, and follow-up and clip delivery logged until they are done.
What changes
Honest fit
For you if: Your guest bookings live in a scramble of mentions, memory, and untracked promises of follow-up and clips.
Skip it if: You need outreach sent for you, contact details found for strangers, or episodes edited and published; prospects come from your own network and past episodes, and you send every message yourself.
Your first run produces: one pipeline record for the next guest you plan to invite: one prospect from your own network or a past episode with the source written in plain words, the outreach drafted and held for your approval and sent by you, the sent date logged, and a prep sheet built from the guest's public work when they agree.
Preview, from the skill itself
“# Podcast Guest Pipeline<br># Podcast Guest Pipeline A guest booking often lives in a scramble: a name mentioned in passing, an outreach note sent from memory, a recording booked with no prep, and clips promised afterwards that no one tracks.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: 0058a5e681ff… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 32.0 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/indie-maker-first-booking/WALKTHROUGH.md, examples/returning-author-second-visit/WALKTHROUGH.md, reference/outreach-drafts-and-approval.md, reference/prep-sheet-from-public-work.md, reference/prospect-sourcing-and-tracking.md, reference/recording-follow-up-and-clips.md, templates/guest-pipeline-record.template.md, templates/guest-prep-sheet.template.md
Worked example, from the package
examples/indie-maker-first-booking/WALKTHROUGH.md
This is a fictional example. The guest, the show, and every name and date in it are made up for this walkthrough, and the figures are internal pipeline figures only.
Learn what a new agent can actually touch, what it sends and where, and how it fails, proven on copies before it ever sees a real record, with the admit or refuse call on record.
What changes
Honest fit
For you if: You are about to let a new agent start live work, and what it can reach, what it sends, and how it fails are known only from its description.
Skip it if: You need the files inside a skill package scanned before they are trusted, access granted, or a faulty agent fixed; vetting reads the package, and this intake tests the running agent on copies and grants nothing itself.
Your first run produces: one intake record for the next agent you were about to let start live work: the proposed role in one line, a small set of copies with four failure probes run, the touch map and the sends map drafted from what was seen, and the admit, admit narrower, or refuse call recorded with its reasons.
Preview, from the skill itself
“# New Agent Intake Checks<br># New Agent Intake Checks A new agent can sound ready long before it is safe to let near real work.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: 629489d46970… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 41.5 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/inbox-triage-agent-for-a-trades-business/WALKTHROUGH.md, examples/supplier-ordering-agent-for-a-small-shop/WALKTHROUGH.md, reference/map-what-it-can-touch.md, reference/map-what-it-sends-and-where.md, reference/prove-failure-behaviour-on-copies.md, reference/the-admit-or-refuse-record.md, templates/intake-checklist.template.md, templates/intake-decision-record.template.md
Worked example, from the package
examples/inbox-triage-agent-for-a-trades-business/WALKTHROUGH.md
A small trades business gets more enquiries than the owner can sort each morning: quote requests, date changes, small job offers, and follow ups on work in hand. A triage agent is offered for the job. The proposed role is one line long: sort copied enquiries, draft replies for the owner, and send nothing. The owner likes the idea and does not like the risk, so the agent goes through intake before it sees a single real enquiry. This is how that intake ran.
Answer for your whole fleet from one page: every agent's job, what it can touch, who answers for it, and when its entry was last checked, kept in one living register.
What changes
Honest fit
For you if: You run several agents set up at different times, and no one place answers how many there are, what each does, what each can touch, and who answers for it.
Skip it if: You need one new agent brought in, watched while it works, or its output judged; onboarding brings one agent in, and this roster is the standing record that holds the fleet together after that.
Your first run produces: one first roster for the fleet as it stands: every agent you believe exists listed from records, one entry each with its job, status, touch list, and named owner, the facts that cannot be sourced written as not yet known and flagged, and the next check date set.
Preview, from the skill itself
“# Agent Fleet Roster<br># Agent Fleet Roster Agents rarely arrive all at once.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: 0e3fab021022… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 39.6 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/first-roster-for-a-three-agent-studio/WALKTHROUGH.md, examples/roster-cleanup-after-a-year-of-growth/WALKTHROUGH.md, reference/keeping-the-roster-current.md, reference/ownership-and-answerability.md, reference/roster-fields.md, reference/what-each-agent-can-touch.md, templates/agent-roster-entry.template.md, templates/fleet-roster.template.md
Worked example, from the package
examples/first-roster-for-a-three-agent-studio/WALKTHROUGH.md
A small studio with five people had grown its agents one at a time over one spring. The owner could name all of them in conversation, but no single page held the fleet, and a new starter asked a plain question nobody could answer from a record: which agent was allowed to write to the shared job board. This is how the first roster was built, in one week, from records.
Turn a procedure only one person knows into a dated runbook entry a second operator can follow at night, checked step by step against the live system.
What changes
Honest fit
For you if: Your important procedures live in one person's memory and scattered notes, so a night incident means calling that person or rebuilding the fix from fragments.
Skip it if: You need instructions for end users of a finished product, a system monitored, or an incident fixed for you; a runbook is for the operator who runs the fix, and user guides are a separate skill.
Your first run produces: one checked runbook entry for the procedure you would least want to explain over the phone at night: the steps listed from the last real record, drafted with its trigger, checks, and stop and escalate point, confirmed against the live system, dated, and added to the index with its review date.
Preview, from the skill itself
“# Runbook Writing<br># Runbook Writing Some procedures only fail when they are needed most.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: 4ba69f557751… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 41.7 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/nightly-backup-failure/WALKTHROUGH.md, examples/web-service-restart/WALKTHROUGH.md, reference/anatomy-of-a-runbook-entry.md, reference/checking-against-the-live-system.md, reference/keeping-runbooks-current.md, reference/writing-operator-steps.md, templates/runbook-entry.template.md, templates/runbook-index-and-review.template.md
Worked example, from the package
examples/nightly-backup-failure/WALKTHROUGH.md
A small studio keeps its project files on one file server. A backup job named Nightly Backup runs at 01:30 each night and writes a copy to a backup volume of 200 GB. One operator, Sam, knew how to handle a failed run. No one else did, and nothing usable was written down. When a run failed while Sam was on leave, the morning cover operator found the failure notice and had to wait three hours for Sam to be reached. The studio decided that procedure would be the first runbook entry. This is how that entry was written.
Catch a dead on arrival build in the studio instead of in front of a user, with one fixed pre-ship list for start, main flows, version stamp, and artifacts, and a written record for every build.
What changes
Honest fit
For you if: You ship builds, and checking is done from memory in a different order each time, so front door faults like a blank first screen ship and ship again.
Skip it if: You need automated tests written, faults fixed, or a release published for you; this is the fixed manual check on the exact artifact, and the ship step stays yours.
Your first run produces: one verification checklist for the product you release most often, and a practice run of the full list on your most recent build: the main flows named, the version match written down, and every check marked passed, failed, blocked, or not run in a complete per build record.
Preview, from the skill itself
“# Build Verification Checks<br># Build Verification Checks Most dead on arrival builds did not fail in a hidden place.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: e0332003a75e… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 39.9 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/mobile-app-blank-screen-catch/WALKTHROUGH.md, examples/service-release-clean-check/WALKTHROUGH.md, reference/recording-a-build-result.md, reference/the-verification-checklist.md, reference/version-and-artifact-checks.md, reference/when-a-check-fails.md, templates/build-verification-record.template.md, templates/verification-checklist.template.md
Worked example, from the package
examples/mobile-app-blank-screen-catch/WALKTHROUGH.md
A small studio builds a field notes app for phones. The app had shipped six earlier versions with checking done from memory. For version 1.8.0 the studio wrote its first fixed checklist and ran it on every build. This is the run that caught a build which would have reached users showing nothing but a blank screen.
Grade every source behind your research before its claims are used, so each claim carries only the weight its source earned, with copies counted once and conflicts named in the open.
What changes
Honest fit
For you if: Your briefs rest on sources that all look alike in the notes, so the weakest one, a copied summary or an undated page, can end up carrying the decision.
Skip it if: You need claims themselves verified as true, sources found and read for you, or a conflict settled; grading weighs the source for a stated use, and claim checking is the Verified Research skill.
Your first run produces: one source register for the next brief you were about to write up: its four to six sources listed with their claims, each graded with its reason and what would change the grade, repeated claims grouped to one origin, and one grade card filled in for the source the brief leans on hardest.
Preview, from the skill itself
“# Research Source Grading<br># Research Source Grading A claim is only as strong as the source behind it, but research notes rarely show that strength.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: 594836028ac4… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 45.7 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/grading-a-vendor-performance-claim/WALKTHROUGH.md, examples/grading-sources-for-a-pricing-brief/WALKTHROUGH.md, reference/conflicts-and-grade-changes.md, reference/dates-publishers-and-provenance.md, reference/primary-and-secondary-sources.md, reference/the-grading-scale.md, templates/source-grade-card.template.md, templates/source-register.template.md
Worked example, from the package
examples/grading-a-vendor-performance-claim/WALKTHROUGH.md
This is a fictional example built for this package. The tool, the maker, and every figure in it are internal to the example, and the figures are consistent within it so the grading moves can be followed end to end.
A library you can find things in again: names, dates, and embedded fields agree under one written rule, with every change logged per file and originals kept untouched until you sign off.
What changes
Honest fit
For you if: Your files carry names and fields that disagree with each other, so a file is hard to find, trust, or tell apart from the one next to it.
Skip it if: You need file contents fixed, damaged files repaired, files converted, or a ruling on what to keep and throw away; this covers names, dates, and metadata only, and nothing here deletes a file.
Your first run produces: one small batch cleaned under a written rule: the batch read as found with every disagreement on paper, a plan with a row per file, the originals set aside untouched, and only the approved changes applied, each written into the per file change log as it lands.
Preview, from the skill itself
“# File Metadata Cleanup<br># File Metadata Cleanup A library goes hard to use in a quiet way.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: fcc26ba71f03… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 49.7 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/client-document-archive/WALKTHROUGH.md, examples/spring-photo-export/WALKTHROUGH.md, reference/approval-originals-and-verification.md, reference/batch-checking-and-holds.md, reference/metadata-fields-and-sources.md, reference/writing-the-cleanup-rule.md, templates/cleanup-plan.template.md, templates/per-file-change-log.template.md
Worked example, from the package
examples/client-document-archive/WALKTHROUGH.md
A small studio finished a cafe fit out for a client, Harbor Lane Cafe, and went to hand over the project archive: estimates, invoices, a schedule, and sign off sheets gathered over the job. The archive held 30 files from the whole job, and it showed. Some files were named with dates, some with working titles, and the titles and authors inside the files often belonged to the blank documents they had started from, not to this job. The client would keep this archive as their record of the work, so a wrong title or author in it would be read as fact for years. The operator wanted the batch cleaned against the studio job sheet, which listed each document on the job, its proper title, and its date. This is how that batch was cleaned. All counts below are from this batch only.
Your first call with a new lead runs from a one-page script built from your own services, so the job gets sized, only deliverable promises get made, and the call ends with one next step both sides heard the same way.
What changes
Honest fit
For you if: Your first calls with new leads are where jobs get missold: a price named before the job is sized, a day offered before the book is checked.
Skip it if: You need leads captured and routed, prices set, or the calls made for you; Lead Intake routes the lead, and this skill scripts the conversation, which you still hold yourself.
Your first run produces: one script for the next new lead already on the books: the services the conversation may offer, five to eight sizing questions in running order, both promise lists in the words you would say out loud, and the one next step checked as deliverable, held for your approval before anything is said.
Preview, from the skill itself
“# First Conversation Scripts<br># First Conversation Scripts The first real conversation with a new lead is where a job is sized, and where it is most often missold.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: ea7805d70840… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 33.6 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/fence-repair-first-call/WALKTHROUGH.md, examples/junk-haul-first-call/WALKTHROUGH.md, reference/building-the-script-from-your-services.md, reference/questions-that-size-the-job.md, reference/the-one-next-step.md, reference/what-can-and-cannot-be-promised.md, templates/conversation-record.template.md, templates/first-conversation-script.template.md
Worked example, from the package
examples/fence-repair-first-call/WALKTHROUGH.md
A fence repair operator takes panel and post work only. No new full fences, and no gate fitting. A new lead calls on Tuesday about a back fence damaged in wind. The operator runs the first conversation from the script built for fence repair leads, approved the week before.
Finds extra work a job record genuinely supports, writes each add-on into the quote as its own line at its price from the operator price list, offers it once in plain words, and records the customer answer either way.
What changes
Honest fit
For you if: You run local service jobs and sometimes see honest extra work next to the booked work, and you want it offered cleanly, priced from your own list, and settled on the record before the job starts.
Skip it if: You want a script to talk a customer into work the job does not call for, or you do not keep a price list and a job record the offer can be checked against.
Your first run produces: one quote checked before it goes out: the nearby work listed, each item tested against the job record, the kept add-ons priced from your price list as their own lines, and the offer drafted in plain words, held for you to send, with the answer to be written on the record the day you hear it.
Preview, from the skill itself
“# Add-On Offers<br># Add-On Offers Extra work is easy to get wrong in two directions.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: 3d0cd3a99ffe… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 34.2 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/deck-repair-stain-and-seal/WALKTHROUGH.md, examples/kitchen-repair-grout-and-caulk/WALKTHROUGH.md, reference/offering-once-and-recording-the-answer.md, reference/price-list-and-what-is-never-offered.md, reference/when-the-record-supports-an-add-on.md, reference/writing-the-add-on-as-its-own-line.md, templates/add-on-offer-record.template.md, templates/quote-with-add-ons.template.md
Worked example, from the package
examples/deck-repair-stain-and-seal/WALKTHROUGH.md
A local repair operator is booked to replace damaged boards on a back deck and reset a loose rail section. On the visit the operator noted the state of the deck finish. That note is what makes one honest add-on possible on this job, and the record is also what stops a second candidate from being offered at all.
A bigger job billed in stages agreed before work starts, with a deposit, midpoints tied to visible milestones, and the balance at completion, each stage recorded as open or paid so money and work stay in step.
What changes
Honest fit
For you if: You run bigger jobs that run past a few days, and payment rests on a deposit and a hope, with midpoints argued from two different memories of how far along the work is.
Skip it if: You need the job priced, a deposit set as a standing term, invoices chased, or money taken for you; Deposit Terms Setup sets the deposit alone, and this schedule drafts and records while you take the payments.
Your first run produces: one schedule for the next bigger job on the books, built before the quote goes out: the total taken from the accepted estimate, split into deposit, midpoints, and balance, each midpoint named by a milestone the customer can see, the stages shown to add to the total, and the stage record opened beside it.
Preview, from the skill itself
“# Progress Payment Schedule<br># Progress Payment Schedule A bigger job paid all at the end leaves the operator carrying weeks of work and materials on trust.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: a68665ea8060… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 36.5 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/kitchen-remodel-three-stages/WALKTHROUGH.md, examples/roof-replacement-milestone-slip/WALKTHROUGH.md, reference/changes-and-balance-at-completion.md, reference/customer-wording-for-stages.md, reference/stage-plan-tied-to-visible-milestones.md, reference/stage-record-open-or-paid.md, templates/progress-payment-schedule.template.md, templates/stage-payment-record.template.md
Worked example, from the package
examples/kitchen-remodel-three-stages/WALKTHROUGH.md
A kitchen fitter is taking out an old kitchen and fitting a new one in the same room: new cabinets, worktops, tiling, and the sink and hob connected. The job will run across two weeks, and the customer does not want to pay the whole total at either end of it. The operator and the customer agree a staged schedule before any work starts.
Every trip your work vehicle makes is logged against the job it served and every fuel stop is recorded from its receipt, so each closed job shows its true road cost and the year's vehicle totals exist as arithmetic, not recollection.
What changes
Honest fit
For you if: You drive a work vehicle from job to job and your job costs never include the road, so a long drive can quietly eat the profit on a small job.
Skip it if: Your vehicle costs are already tracked per job in another system you trust, or you do not run jobs from a vehicle at all.
Your first run produces: one log set up for the vehicle that does most of the job driving, with today logged as the first day: the vehicle header and odometer reading written, today trips each naming a job with their readings, today fuel stops recorded from their receipts, and the day miles adding up to the difference between the first and last readings.
Preview, from the skill itself
“# Fuel and Mileage Log<br># Fuel and Mileage Log A job can look profitable on the invoice and still lose money on the road.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: 7d7b3d60f65d… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 33.7 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/fence-repair-two-jobs-one-van/WALKTHROUGH.md, examples/firewood-delivery-week/WALKTHROUGH.md, reference/closing-the-month-and-the-year.md, reference/logging-a-trip-against-a-job.md, reference/recording-a-fuel-stop.md, reference/turning-the-log-into-a-job-cost.md, templates/job-and-year-totals.template.md, templates/trip-and-fuel-entry.template.md
Worked example, from the package
examples/fence-repair-two-jobs-one-van/WALKTHROUGH.md
An operator repairs fences from one van. On Monday the van serves two jobs: a fence repair that runs into Tuesday, and a small gate hinge fix across town. On Tuesday the operator goes back to the fence repair alone to finish it. One fuel stop falls on Tuesday morning.
The crew starts knowing the site instead of discovering it, because one recorded walk before work names hazards across access, overhead, ground, and weather exposure, gives each a written plan, and files the photos against the job.
What changes
Honest fit
For you if: You run small crews on houses and yards where the site changes from job to job, and the first hour keeps being spent on finds the start of the day could have caught.
Skip it if: Your work is in one fixed place you walk daily under a standing plan of your own, or you need a formal inspection or a signed clearance, which this record is not.
Your first run produces: one walk record for the next job, made before any tools come out: the site walked in the set order from the job record, each hazard named with its place and given a plan and an owner, the photos filed against the job by number, and the crew brief drafted from the record, waiting for your word.
Preview, from the skill itself
“# Site Safety Walk<br># Site Safety Walk Most site trouble is not a surprise.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: e45376abb654… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 38.8 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/brush-clearing-deck-teardown/WALKTHROUGH.md, examples/roof-valley-repair-steep-lot/WALKTHROUGH.md, reference/access-and-overhead-hazards.md, reference/ground-and-weather-hazards.md, reference/photos-filed-against-the-job.md, reference/plan-for-each-hazard.md, templates/crew-start-brief.template.md, templates/safety-walk-record.template.md
Worked example, from the package
examples/brush-clearing-deck-teardown/WALKTHROUGH.md
An operator and three crew are set to clear brush along a fence line and take down an old raised deck at the back of a sloped yard. The work needs saw work in the brush, boards carried up the yard, and waste carried down to a trailer at the street. The day is set to be hot. No work starts until the walk is done.
Every subcontractor payment tied to the agreed stage it covers, partials shown as partial, and the balance owed known from one record at any point.
What changes
Honest fit
For you if: You pay one or more subcontractors against written agreements and want paid, partial, and still owed to be known from one record at any point, without rebuilding it from bank lines and memory.
Skip it if: You have no written agreement with the sub to pay against, or one person pays a single fixed sum on completion every time with nothing staged, partial, or extra in between; in that case a single paid line on the job record already does the whole job.
Your first run produces: one payment record for the next subcontractor agreement already on the books: the agreed total written exactly as agreed and split into its stages, the stage amounts checked against the total, any payments already made entered and tied to their stage or flagged, and the balance worked from the lines.
Preview, from the skill itself
“# Subcontractor Payment Records<br># Subcontractor Payment Records A subcontractor can do good work and still leave a money argument behind.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: d2a4535932a8… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 35.2 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/deck-framing-staged-payments/WALKTHROUGH.md, examples/fence-build-agreed-extra/WALKTHROUGH.md, reference/agreement-total-and-stages.md, reference/extras-holdbacks-and-closing-the-record.md, reference/partial-payments-and-balance-owed.md, reference/recording-a-payment-as-it-happens.md, templates/payment-ledger.template.md, templates/payment-summary-for-the-sub.template.md
Worked example, from the package
examples/deck-framing-staged-payments/WALKTHROUGH.md
A decking operator has taken a full deck job and brings in Marcus, a carpenter, to build the frame while the operator handles the decking boards, steps, and finish. The frame money is agreed in writing before Marcus starts, and this walkthrough follows only that money: the agreed total, its stages, two payments, and the balance still owed part way through the job.
Settle any questioned agent decision from one entry instead of from memory, and catch a repeating fault in the log before it becomes damage: every consequential call is recorded at the time with what the agent was given, the rule or record it followed, and what came of it.
What changes
Honest fit
For you if: Your agents approve, refuse, route, spend, or promise in your name, and a questioned call would today be settled from memory and from records that have since changed.
Skip it if: You need the decision made or judged for you, or one incident worked in depth; this log is the standing record those readings work from, and it makes no decision, widens no gate, and fixes no fault itself.
Your first run produces: one decision log started for one agent: the log opened with its scope line naming the agent and the two or three kinds of call that count as consequential, today's consequential calls entered one per call as they happen, with what each was given, the rule or record it followed, the decision, and the gate the action passed, and the outcome check and first pattern review dates set at the top of the log.
Preview, from the skill itself
“# Agent Decision Log<br># Agent Decision Log An agent makes calls all through the day.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: c5e51c8cf103… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 47.1 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/a-pattern-review-that-changed-an-ordering-rule/WALKTHROUGH.md, examples/a-refund-call-traced-three-weeks-later/WALKTHROUGH.md, reference/decision-entry-fields.md, reference/keeping-and-correcting-the-log.md, reference/reading-the-log-for-patterns.md, reference/what-counts-as-consequential.md, templates/decision-log-entry.template.md, templates/decision-log.template.md
Worked example, from the package
examples/a-pattern-review-that-changed-an-ordering-rule/WALKTHROUGH.md
A repair business ran an ordering and dispatch helper that could approve small parts orders inside a standing grant, pick a supplier where more than one stood, pick a delivery speed, and move job slots when parts ran late. Each of those calls sat inside the scope line of the business's decision log, which the office manager kept. Entries were numbered in one running order. This is the monthly pattern review for June, what it found, and what the next review showed.
Know how much more work your fleet can take, agent by agent, from one record of load, queue growth, and the load where quality starts dropping, before a peak or a growing backlog forces the answer.
What changes
Honest fit
For you if: You run several agents and cannot say from a record which one runs out of room first, or how much more work the fleet can take before quality starts dropping.
Skip it if: You need work taken in, turned away, or moved between agents; this record informs those calls and makes none of them, and it reads the quality checks your fleet already runs rather than judging quality itself.
Your first run produces: one capacity record for the fleet as it stands: one entry per agent for the same recent period, with what counts as one unit of work for that agent, where the count comes from, arrivals and completions, the queue at the start and at the end, and the peak inside the period, then per agent the queue growth and a first reading of sustainable load and headroom, with gaps flagged by name and the next check date set.
Preview, from the skill itself
“# Agent Capacity Planning<br># Agent Capacity Planning A fleet can look busy and healthy right up to the week it falls behind.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: 55781fd1f17e… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 47.2 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/four-agent-fleet-before-the-holiday-peak/WALKTHROUGH.md, examples/weekly-check-that-caught-a-growing-queue/WALKTHROUGH.md, reference/capacity-terms-and-counts.md, reference/finding-where-quality-drops.md, reference/queue-growth-and-backlog.md, reference/reading-per-agent-load.md, templates/agent-load-entry.template.md, templates/capacity-record.template.md
Worked example, from the package
examples/four-agent-fleet-before-the-holiday-peak/WALKTHROUGH.md
A small studio ran four agents for customer work and expected its busiest weeks of the year in late November. The operator had a forecast from the advance orders sheet: about 30 percent more work than the current level, across all four kinds of work. She wanted to know, from the record and not from hope, whether the fleet could take that rise, and where it would run out of room first. This is how the first capacity record was built and read, over one ten working day period.
Every secret on one written rotation schedule, by label and location only: its age, its owner, and its next rotation date, with the reminder drafted ahead of the date and each rotation logged, so no secret ages in silence.
What changes
Honest fit
For you if: You hold keys, tokens, or passwords across tools and cannot say from one page how old each one is, who owns it, or when it is next due to rotate.
Skip it if: You need the rotation done for you, or rules for how a secret value may be stored and handled; this schedule rules the calendar only, holds no value, and each rotation stays in your hands in the storage tool.
Your first run produces: one rotation schedule built from the records as they stand: every secret listed one line each, by label and location only, each line filled with its owner, its last rotated date read from a record, and an interval set for it, the age and the next date worked out from the dates rather than from memory, secrets of unknown age flagged and planned for a first rotation soon, and reminders drafted for every date due in the next thirty days.
Preview, from the skill itself
“# Secret Rotation Schedule<br># Secret Rotation Schedule Most secrets are rotated on the day something goes wrong, or on the day someone happens to remember.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: 4480723cf40b… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 41.0 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/catching-up-an-overdue-schedule/WALKTHROUGH.md, examples/first-schedule-for-a-solo-studio/WALKTHROUGH.md, reference/choosing-rotation-intervals.md, reference/logging-a-rotation.md, reference/reminders-and-the-rotation-day.md, reference/schedule-fields.md, templates/rotation-log-entry.template.md, templates/secret-rotation-schedule.template.md
Worked example, from the package
examples/catching-up-an-overdue-schedule/WALKTHROUGH.md
A three person team, Ana, Ben, and Cleo, kept a rotation schedule for the twelve secrets behind their booking service. The schedule had been built the previous November and kept well until April. Then the keeper changed: Ben took the schedule over in May, the May review was missed in a launch week, and no reminders went out for six weeks. The catch up review was held on Monday 8 June 2026, with the schedule and the log side by side.
Get every published page checked against one fixed accessibility list, with findings logged per page and fixes drafted for your approval, so weak alt text, broken heading order, vague link names, and failing contrast are found and settled from values instead of guessed at by eye.
What changes
Honest fit
For you if: You publish pages and cannot say which ones hold images without alt text, broken heading order, vague link names, or text that fails on its contrast values.
Skip it if: You need keyboard use, focus order, forms, motion, or reading software behaviour checked, or a claim that a page meets every rule or law that can apply to it; this list covers four heads only, and fixes are drafted for your approval, never published by the skill.
Your first run produces: one check log for the published pages as they stand: every page in scope listed one line each and settled against the live site, then one page at a time worked through the fixed list, every image for alt text, every heading in order, every link name in full, and every text and background pair from the design values in use, each finding written on its page entry and each fix drafted on the fix draft sheet for your approval, with the next check date set.
Preview, from the skill itself
“# Page Accessibility Checks<br># Page Accessibility Checks A published page is edited many times by many hands.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: 131d4a167b36… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 44.7 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/event-listing-and-guide-pages-for-a-community-site/WALKTHROUGH.md, examples/home-and-services-pages-for-a-small-studio/WALKTHROUGH.md, reference/alt-text-and-image-honesty.md, reference/contrast-from-design-values.md, reference/headings-and-link-names.md, reference/the-fixed-check-list.md, templates/fix-draft-sheet.template.md, templates/page-check-log.template.md
Worked example, from the package
examples/event-listing-and-guide-pages-for-a-community-site/WALKTHROUGH.md
A volunteer run community site listed local events and held guides for visitors. Five pages were published: an events home page, a listing for the current month, a venue guide, a tickets guide, and a page about the group. Pages were edited by whoever had time, and the site had never had a single check. This is how the first check was run, over two weeks of short sittings, from the pages and the design record as they stood.
Ship every release knowing, from one ledger rechecked against the build, what each dependency carries and what its license asks, so no obligation goes out unrecorded.
What changes
Honest fit
For you if: You ship software and cannot say from one page what each dependency carries or what its license asks, so an obligation can go out unrecorded with a release.
Skip it if: You need legal advice, a ruling on whether two licenses can sit together in one product, or dependencies chosen for you; the ledger records obligations and makes none of those calls.
Your first run produces: one ledger for the next release, built from the build as it stands: every dependency you believe ships listed one line each from the manifest, the lock file, the build list, and the bundled items, then one entry per dependency with its name, the version shipped, the license for that version, what that license asks, where the license text was read, and the date those facts were checked at their source, with gaps flagged as not yet known and the next recheck set for the next release.
Preview, from the skill itself
“# Dependency License Ledger<br># Dependency License Ledger A product rarely ships alone.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: 5336fc2d18b8… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 41.6 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/first-ledger-for-a-shipping-app/WALKTHROUGH.md, examples/release-recheck-that-caught-a-license-change/WALKTHROUGH.md, reference/keeping-the-ledger-current.md, reference/ledger-fields.md, reference/reading-a-license-for-the-ledger.md, reference/what-each-license-asks.md, templates/dependency-license-entry.template.md, templates/license-ledger.template.md
Worked example, from the package
examples/first-ledger-for-a-shipping-app/WALKTHROUGH.md
A small studio was one week from the first release of a desktop invoicing tool for solo shops. The tool had been built over one season, with libraries added as each feature landed. The owner could name most of them in conversation, but no single page said what the installer carried, under which licenses, or what those licenses asked for. With the release close, she asked for one true ledger before anything went out. This is how the first ledger was built, over three days, from records.
Find any archived file by what it is, with a searchable index built from each file's subject, date, parties, and type and kept current as new files arrive.
What changes
Honest fit
For you if: You keep an archive of contracts, job files, or records, and finding one of them means recalling its folder or its file name, or asking whoever filed it.
Skip it if: You need files moved or renamed, a file's contents summed up or judged, or search inside the full text of files; the index points to each file as it stands and changes neither its place nor its name.
Your first run produces: one index for the archive as it stands: a fresh inventory of every file in scope, one line each, from the archive itself, then one entry per file with its subject in plain words, the date on the file, the parties it names, its type, and the place it sits, each field read at its source and any field that cannot be found flagged as not yet known, finished by testing the index with real searches and setting the intake habit and the next check date.
Preview, from the skill itself
“# Archive Search Index<br># Archive Search Index An archive can hold a file safely and still lose it in practice.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: d401f5994c7e… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 43.6 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/first-index-for-a-contracts-archive/WALKTHROUGH.md, examples/keeping-a-job-archive-searchable/WALKTHROUGH.md, reference/index-fields.md, reference/keeping-the-index-current.md, reference/reading-fields-from-the-record.md, reference/searching-and-checking-the-index.md, templates/archive-index.template.md, templates/index-entry.template.md
Worked example, from the package
examples/first-index-for-a-contracts-archive/WALKTHROUGH.md
A small studio kept its client papers in one archive built up over five years: contracts, letters about them, invoices, and short reports on finished work. The papers were safe, in nine folders by year and client, but finding one meant recalling the year it was filed and the client folder it sat in. When a past client asked for a copy of a letter about a change to their contract, two people spent an afternoon opening folders before it turned up in a different year from the one both of them recalled. This is how the first index was built, in one week, from the files themselves.
Your seasonal rushes get planned in the quiet months: last year's own job records become a season map with an opening date and an honest slot count per season, plus an ordered waitlist kept for every season that fills.
What changes
Honest fit
For you if: Your work comes in seasonal rushes, and the rush is survived in the busy months each year instead of planned in the quiet ones, when you can still choose.
Skip it if: You need offers written, prices set, jobs booked, or customers contacted; this calendar is the timing those rest on and does none of them, and a new service with no past year gets a named gap on the calendar, not a guessed season.
Your first run produces: one whole year map built in one sitting from the last full year of job records: every seasonal job tallied by the month it was done and the service it was, the months grouped into the seasons the tallies actually show, each season's opening date set ahead of when its first enquiries arrived last year, each season's slots counted from the crew and days you can truly give it, and one empty waitlist started per season, ready for names.
Preview, from the skill itself
“# Seasonal Booking Calendar<br># Seasonal Booking Calendar Every local service year has the same rushes in the same places.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: 6acb317e2698… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 36.0 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/exterior-washing-gutters-lights/WALKTHROUGH.md, examples/leaf-and-yard-spring-fall/WALKTHROUGH.md, reference/mapping-demand-to-seasons.md, reference/opening-dates-and-capacity.md, reference/reading-past-job-records.md, reference/waitlists-that-hold.md, templates/seasonal-calendar.template.md, templates/waitlist-sheet.template.md
Worked example, from the package
examples/exterior-washing-gutters-lights/WALKTHROUGH.md
A home exterior operator with two crews, doing pressure washing, gutter clearing, and holiday light fitting and removal. Two of its seasons share November, and its tightest season is its shortest one. In January 2026 the operator builds the year map from the full 2025 job records.
Your estimates stop quoting last season's prices: the materials you buy most are rechecked on a schedule from real listings at the exact quantity, every check dated and sourced, changes logged, and anything that cannot be sourced carried as a flagged placeholder, never a made-up figure.
What changes
Honest fit
For you if: You write estimates, and the material prices inside them are last season's figures, remembered or copied from quote to quote, with no date and no source behind them.
Skip it if: You need materials bought, orders placed, or selling prices, margins, or quote totals set; this skill only keeps the price record, and it never promises that a listed price will still hold at the counter.
Your first run produces: one watch list and its first check in the same sitting: the materials that show up again and again named from your purchase notes and most recent estimates, each with the exact quantity its price will be held in, then the first check run, each item priced from a real listing for that exact quantity and logged with the listing named as the source and today's date as the check date, any item with no such listing entered as a flagged placeholder with the reason and no figure, and the check schedule written at the top with the next check date named.
Preview, from the skill itself
“# Materials Price Watch<br># Materials Price Watch An estimate is only as good as the prices inside it, and materials prices do not sit still.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: 45041066fa7f… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 34.2 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/fence-and-deck-may-check/WALKTHROUGH.md, examples/roofing-and-lumber-march-check/WALKTHROUGH.md, reference/building-the-watch-list.md, reference/checking-prices-by-exact-quantity.md, reference/marking-figures-for-estimates.md, reference/reading-changes-and-flagging-gaps.md, templates/estimate-price-note.template.md, templates/price-watch-log.template.md
Worked example, from the package
examples/fence-and-deck-may-check/WALKTHROUGH.md
A local service operator who builds and repairs fences and small decks. Five materials sit on the watch list from the purchase records. The list is checked every two weeks. This is the check run on 5 May 2026, against the last check on 21 April 2026, which is 14 days earlier, as the schedule sets.
Pay and true job labor cost come off one sheet instead of memory: every crew member's hours are entered per job on the same day, with totals per job and per week worked in the open.
What changes
Honest fit
For you if: You run a small crew of your own, pay them by the hour, and want pay and true job labor cost to come from one record you can check, not from a rebuilt week.
Skip it if: You run alone with no crew hours to track, or every hand on your jobs is an outside sub paid against an agreement; in that case the subcontractor payment record already does the whole job.
Your first run produces: one complete week sheet for the current week, or the last full week the crew worked: the crew listed by name with the agreed rate for each person exactly as the record holds it, the jobs that week listed from the job records, the hours entered day by day, one line per person per job per day, days off written as days off in words and gaps left named, then the totals per person, per job, and per week worked from the lines and checked three ways, with each pay figure tied to a person, a total, and a rate.
Preview, from the skill itself
“# Crew Timesheet Records<br># Crew Timesheet Records At the end of a busy week, hours are the easiest thing to lose and the hardest thing to rebuild.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: cbf0822e8877… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 38.3 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/three-person-crew-one-job-week/WALKTHROUGH.md, examples/two-jobs-split-week-corrections/WALKTHROUGH.md, reference/corrections-gaps-and-missing-days.md, reference/entering-hours-the-same-day.md, reference/rates-pay-and-job-cost.md, reference/totals-per-job-and-per-week.md, templates/job-hours-summary.template.md, templates/weekly-timesheet.template.md
Worked example, from the package
examples/three-person-crew-one-job-week/WALKTHROUGH.md
A local service operator with a crew of three on one job for a full week, Monday to Friday: a barn repair and cleanout in Oakdale. Every hour was entered on the day it was worked, and the week was closed from that one sheet.
Every work trip starts from a signed record instead of a shrug: one standard check of lights, tires, hitch and coupler, and load secured, logged per trip against the job, so a roadside failure or a load question has a record behind it.
What changes
Honest fit
For you if: You run work trips by truck and trailer, and a roadside failure or a question about a load would today be met by recollection instead of a record.
Skip it if: You need a roadworthiness certificate, a mechanic's repair, or a ruling on legal weight or a vehicle rating; a signed sheet is none of those, and the decision on every sheet is yours, recorded, not made for you.
Your first run produces: one complete sheet for the next real work trip, in about twenty minutes before it leaves: the top block filled from the records with the job the trip serves, the vehicle, the trailer if one is going, and who is running the check, then the rig walked once, slowly, in the sheet's order, with the tire pressures the vehicle maker states and the ball size the coupler is marked for written on the sheet itself, all 15 items marked, and the result and decision recorded and the sheet filed against that job's record.
Preview, from the skill itself
“# Vehicle and Trailer Checks<br># Vehicle and Trailer Checks A trailer light that was dark all week, a tire that was soft for three trips, a coupler that was never quite latched: none of these announce themselves at the yard.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: 74d2fd5195cf… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 34.6 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/clean-check-haul-to-job/WALKTHROUGH.md, examples/failed-light-held-at-yard/WALKTHROUGH.md, reference/hitch-coupler-and-safety-chains.md, reference/lights-tires-and-running-gear.md, reference/load-secured-and-ready-to-roll.md, reference/when-a-check-fails.md, templates/check-log.template.md, templates/trip-check-sheet.template.md
Worked example, from the package
examples/clean-check-haul-to-job/WALKTHROUGH.md
A solo operator, Dale, running a pickup and a single axle utility trailer to a fence repair job in Harriman on Tuesday 14 April 2026. One trip, one sheet, every item passing.
Hand-typed records earn trust before later work stands on them: every field gets a second pass against its source, mismatches are listed by field, corrections come from the source only, and anything that cannot be read is flagged, never guessed.
What changes
Honest fit
For you if: Later work stands on records typed by hand from notes, cards, or forms, and nothing says which of them were ever checked against their sources.
Skip it if: You need the source itself judged right, a faint value settled between two readings, or records checked that were produced by a system transfer with no hand typing in them; live records change only after you read the sheet and approve it.
Your first run produces: one verification sheet for one small set of hand typed records, twenty to fifty, taken with the source documents they were typed from: the records and the sources laid side by side in the same order, the fields to be checked named at the top of the sheet, the set worked record by record with each field compared to its source, each mismatch logged by field as it is found, each correction copied from the source only, and any field the source cannot settle flagged with its reason, the counts reconciled at the end.
Preview, from the skill itself
“# Data Entry Verification<br># Data Entry Verification Records typed in by hand carry the slips of the typing: a digit turned around, a date copied from the line above, a name spelled the way it sounded.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: 22f34dd538bf… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 30.5 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/forty-eight-delivery-notes-stock-sheet/WALKTHROUGH.md, examples/one-hundred-twenty-signup-cards-contact-list/WALKTHROUGH.md, reference/comparing-field-by-field.md, reference/flagging-unreadable-fields.md, reference/mismatch-log-and-corrections.md, reference/source-is-the-authority.md, templates/mismatch-log.template.md, templates/verification-sheet.template.md
Worked example, from the package
examples/forty-eight-delivery-notes-stock-sheet/WALKTHROUGH.md
A small supplier business typed a week of paper delivery notes into its stock sheet by hand, 48 records in all, before the sheet was used to place the next orders. This pass checks those 48 records against the notes themselves before any order stands on them.
One name means one entry: duplicates are merged with every merge on record, stale entries and conflicting details are flagged for your call with both values kept, and nothing is deleted in silence.
What changes
Honest fit
For you if: Your contact list holds the same people and businesses more than once, with their details split across the copies, and stale entries hiding among the live ones.
Skip it if: You need entries removed, a conflict between two values settled, or a number checked to see if it still rings; flagged entries wait for your call, both values are kept, and removal is your act.
Your first run produces: one full hygiene pass on the working contact list, or one part of it with at least twenty entries in it: the entries counted and the count written down, a copy sorted by name and by phone number, the groups that look like the same person or business checked group by group, the groups that pass merged with one merge log line per merge, any field where a group disagrees flagged, then the merged list worked for stale entries, each flagged with its reason, and the canonical list and hygiene report held for your approval, with no entry removed.
Preview, from the skill itself
“# Contact List Hygiene<br># Contact List Hygiene A working contact list drifts.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: e4f711398257… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 39.8 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/small-contractor-list-cleanup/WALKTHROUGH.md, examples/supplier-and-customer-list-cleanup/WALKTHROUGH.md, reference/building-the-canonical-entry.md, reference/finding-duplicates-without-guessing.md, reference/merging-and-the-merge-record.md, reference/stale-entries-and-conflict-flags.md, templates/hygiene-report.template.md, templates/merge-log.template.md
Worked example, from the package
examples/small-contractor-list-cleanup/WALKTHROUGH.md
A one crew contractor keeps a working contact list of customers, suppliers, and rental yards in one file. The list has grown by entry after entry for two years, with no pass ever made over it. The operator sets a stale line for this pass: no contact noted in the last 18 months counts as stale evidence where an entry holds contact notes at all.
A closed incident gets one true account instead of two memories: the timeline built from the record and the logs as they actually ran, what held and what failed in plain words, and every fix carried out with an owner, a date, and the check that will show it landed.
What changes
Honest fit
For you if: A service incident has closed and you need one true account of what happened and what will change, built from the record and the logs rather than from memory.
Skip it if: The incident is still live and you need updates drafted while it runs, or the incident was an agent's behaviour to review; this skill starts only once a service incident has closed, and it fixes nothing itself.
Your first run produces: one short postmortem for the most recent closed incident you have, however small: the sources gathered first, the incident record, the logs for its window, and the messages that went out while it was live, with any missing source written down as a gap, the timeline built one line per event with its time and its source, what held and what failed sorted against those lines, two fixes drafted with an owner, a date, and a check each, and the whole draft read back against the sources line by line and held for your approval.
Preview, from the skill itself
“# Incident Postmortem Writing<br># Incident Postmortem Writing An incident closes.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: 1be041173f08… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 55.1 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/checkout-timeout-postmortem.md, examples/upload-limit-postmortem.md, reference/building-the-timeline.md, reference/fixes-with-owners-and-dates.md, reference/sources-gaps-and-plain-words.md, reference/what-held-and-what-failed.md, templates/fix-list.template.md, templates/postmortem.template.md
Worked example, from the package
examples/checkout-timeout-postmortem.md
A small online craft shop, run by a two person studio, sells through a hosted checkout. On 14 May the payment step of that checkout began timing out for shoppers. The incident closed the same morning. This is the postmortem as it was written the next day, 15 May, from the incident record and the checkout log: how the sources were gathered, how the timeline was built, how held and failed were sorted, and the fixes that came out of it, with the check pass at the end.
Work stops piling onto one agent while another sits idle: the overloaded and the underused are read from the capacity record, each move is proposed with its reason and approved before it happens, and the result is checked against what the move promised.
What changes
Honest fit
For you if: Work piles onto one agent while another sits idle, and you hold a capacity record the moves can be read from.
Skip it if: There is no capacity record yet; this skill acts on that record and does not rebuild it, and it never moves work or changes routing on its own.
Your first run produces: one balance worked from reading to check, on the fleet as the capacity record shows it now and the one queue that is growing: the record read for which agent is overloaded and which agents have room below their limits, one move proposed with the work, the count, both agents, and the reason in figures, the move carried out only after your approval is noted, counted on both sides, and a check date set on the plan.
Preview, from the skill itself
“# Agent Workload Balancing<br># Agent Workload Balancing A fleet rarely falls behind evenly.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: e1ccd049fa31… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 42.6 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/overloaded-lookup-agent-relieved-by-two-moves/WALKTHROUGH.md, examples/peak-week-rebalance-that-held-quality/WALKTHROUGH.md, reference/balance-terms-and-counts.md, reference/checking-a-move-after-it-happens.md, reference/proposing-and-approving-a-move.md, reference/reading-the-capacity-record-for-moves.md, templates/balance-plan.template.md, templates/move-proposal.template.md
Worked example, from the package
examples/overloaded-lookup-agent-relieved-by-two-moves/WALKTHROUGH.md
A small operation ran three agents that all did the same kind of work: order lookups. One unit of work was one lookup completed, counted from the lookup log, with arrivals counted at entry to the same queue. Each queue was counted at close of day. The three agents were called Ari, Bea, and Cato in the fleet records. This is how one growing queue was read from the capacity record, moved under three approved proposals, and checked a week later.
A third-party skill your fleet depends on never changes under you unnoticed: every package sits in a pinned register, checked against its source on a schedule, and each update is reviewed file by file against the pinned copy before anyone adopts it.
What changes
Honest fit
For you if: Your fleet depends on skills other people wrote, and an update to one of them could change its behaviour or widen its reach without you noticing.
Skip it if: You need a package read once before the fleet takes it on at all; that first read is vetting, and this watch begins after a package is already in the fleet.
Your first run produces: the register built for the fleet as it stands and one real check run before you stop: every third-party skill the fleet depends on listed from the installed copies with its source, pinned version, and reach, each entry checked against its source listing for an update waiting, one full review worked on the entry with an update or the one the fleet leans on hardest, and the watch schedule set at the top of the register with the date of the next pass.
Preview, from the skill itself
“# Third-Party Skill Watch<br># Third-Party Skill Watch A fleet rarely runs only what its operator wrote.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: 53aac98f58ca… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 50.2 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/a-small-patch-reviewed-and-adopted.md, examples/an-update-held-for-a-wider-reach.md, reference/holds-deferrals-and-retiring.md, reference/reviewing-an-update-before-adoption.md, reference/the-dependency-register.md, reference/watch-cadence-and-sources.md, templates/dependency-register.template.md, templates/update-review.template.md
Worked example, from the package
examples/a-small-patch-reviewed-and-adopted.md
An illustrative internal example. All package names and versions in this walkthrough are invented for the example. They are not claims about real packages or real publishers.
The build that ships is the build that was checked: one named candidate is frozen with what is in and what is held out both written down, and a late change gets in only through a written exception the operator approves.
What changes
Honest fit
For you if: A release is close and the candidate keeps moving in its last days, so what ships may not be what was checked.
Skip it if: You need the gate checks themselves run, or the changelog written; this skill holds the candidate still while those run and decides nothing about what deserves to ship.
Your first run produces: one freeze for the next real release on your calendar, however small: the freeze record opened with the product, version, candidate build, and window dates, every item on the change list sorted as in the candidate or held out with a reason and a next home, both lists counted and the opening count reconciled against the change list total, the checks for the window named on the record, and the exception route written down where your team will see it.
Preview, from the skill itself
“# Release Candidate Freeze<br># Release Candidate Freeze In the last days before a release, the candidate keeps moving.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: 11297c65d88e… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 49.3 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/freeze-with-one-admitted-and-one-held-out.md, examples/freeze-with-one-approved-late-fix.md, reference/closing-and-lifting-the-freeze.md, reference/freeze-record-fields.md, reference/the-exception-route.md, reference/what-counts-as-in-or-held-out.md, templates/freeze-record.template.md, templates/late-change-exception.template.md
Worked example, from the package
examples/freeze-with-one-admitted-and-one-held-out.md
A three person team ships Trail Log, a phone app for recording hikes. Release 1.9.0 was planned for Friday 12 June 2026. The founder is the operator, the team's release lead is the keeper, and one builder looks after the app's startup and record screens. The team had run two freezes before this one. This window is the first where a late change was refused, and the refusal is the part of the record the team now points new joiners at.
A screenshot kept as evidence still means something a month later: every capture is logged at the time with what it shows, when it was taken, and which build or record it belongs to, and filed against the record it supports.
What changes
Honest fit
For you if: Screenshots or recordings are kept as evidence for bug reports or job records, and a month later nobody can say what one shows, when it was taken, or which build it belongs to.
Skip it if: You need photographs of physical work on site, or a ruling on whether a claim is true; this record holds what a frame shows, and it widens nothing past the frame.
Your first run produces: the record started for one kind of work where pictures are already taken and later argued about: the index opened with its scope line naming which records captures are filed against, the next real case inside that scope logged capture by capture as taken with what each shows, when it was taken, and the build or record it belongs to, each original filed under the naming method, and the set read back as a stranger would read it before the check schedule is set.
Preview, from the skill itself
“# Visual Evidence Records<br># Visual Evidence Records A screenshot is easy to take and hard to trust later.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: e06b4ea2893b… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 48.4 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/bug-report-checkout-failure/WALKTHROUGH.md, examples/job-record-delivery-dispute/WALKTHROUGH.md, reference/capture-log-fields.md, reference/checking-and-correcting-records.md, reference/naming-and-filing-captures.md, reference/what-makes-a-capture-trustworthy.md, templates/capture-log-entry.template.md, templates/evidence-index.template.md
Worked example, from the package
examples/bug-report-checkout-failure/WALKTHROUGH.md
A small online craft shop ran its shop app on build 3.2.0. On the morning of 4 March a tester working a routine checkout found the total on the checkout screen did not match the cart. The shop kept an evidence index for bug reports, kept by the shop owner. Captures for a bug were filed against the bug report they supported, with one entry per capture. This is how the set for one bug was built, and how it settled the question asked about it fifteen days later.
A pile of files changes format without losing any in silence: the batch runs under a written plan the operator approved, every failure is listed by name, the counts in and out are reconciled, and the originals stay untouched until the batch is signed off.
What changes
Honest fit
For you if: A pile of files has to change format, and losing some of them in silence, or losing figures inside them, would only surface when someone needs the file.
Skip it if: You need a damaged file repaired, a format chosen for you, or a promise that a converted file is the same as its original; the plan names what may change, and damaged files go on the failure list for your call.
Your first run produces: one small batch converted in one sitting, twenty to thirty files of one kind from one place: the batch counted and listed by name, the plan written and approved before any file moves with the formats, the tool, and what must survive the move named, the run worked in plan order with each outcome written into the batch record as it lands, every failure listed by name, the counts reconciled, and a spread of converted files spot checked against their originals while the originals sit untouched.
Preview, from the skill itself
“# Bulk File Conversion<br># Bulk File Conversion A batch of files has to change format.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: bf36b5f5c12b… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 46.6 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/course-handouts-doc-to-pdf/WALKTHROUGH.md, examples/supplier-invoices-pdf-to-sheet/WALKTHROUGH.md, reference/failures-spot-checks-and-reconciliation.md, reference/formats-and-what-conversion-can-lose.md, reference/originals-approval-and-handover.md, reference/writing-the-conversion-plan.md, templates/batch-record.template.md, templates/conversion-plan.template.md
Worked example, from the package
examples/course-handouts-doc-to-pdf/WALKTHROUGH.md
A small training provider held its course handouts as word processor files, most of them written years earlier in an old format the current tools read poorly. The handouts needed to go to learners as PDF files, one PDF per handout, with the pages, the tables, and the diagrams intact. The operator named the tool, approved the plan, and held the originals decision. This is how the batch ran in two passes and still reconciled to one count.
A lapsed quote never quietly comes back at a stale price: every quote carries an honest expiry date, and when one lapses the renewal is rebuilt line by line from your current price record and held for your approval, with the new figure and the change shown in the open.
What changes
Honest fit
For you if: You send quotes that lapse in silence, and a customer can come back weeks later pointing at a stale total.
Skip it if: The quote still stands and you want a reply chased; this skill starts only once the date has passed, and it sets no prices and sends nothing itself.
Your first run produces: one lapsed quote handled from a clean record, in about thirty minutes: the oldest quote past its expiry date with no answer read as sent, the channels checked for any late reply, the quote entered in the expiry log and closed as expired with the lapse date, each renewal line rebuilt from the current price record, and one renewal draft held for you with the new total worked in the open, the change against the old total shown, and a new expiry date written down.
Preview, from the skill itself
“# Quote Expiry and Renewal<br># Quote Expiry and Renewal A quote is a price held for a short time, not a price held for ever.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: ca58f79adbed… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 34.7 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/fence-repair-renewal-after-a-rise/WALKTHROUGH.md, examples/porch-step-renewal-that-closed/WALKTHROUGH.md, reference/closing-and-recording-expired-quotes.md, reference/rebuilding-the-renewal-from-current-prices.md, reference/setting-an-honest-expiry-date.md, reference/writing-the-renewal-note.md, templates/quote-expiry-log.template.md, templates/renewed-quote.template.md
Worked example, from the package
examples/fence-repair-renewal-after-a-rise/WALKTHROUGH.md
A local service operator who repairs fences and gates. A customer, Dale Harper, asked for a price to repair a side fence: four posts to reset, a run of boards to replace, and the ground made good. This example follows the quote from its expiry date, through its lapse in silence, to a renewal rebuilt from current prices and accepted by the customer.
A cancelled job ends as a finished record instead of a standing dispute: the deposit is settled exactly as the written terms say, every materials line is accounted for, the refund is worked in the open, and the customer gets a plain way back.
What changes
Honest fit
For you if: Jobs cancel after deposits are taken and materials are bought, and the settlement today rests on memory and a phone call.
Skip it if: There are no written cancellation terms on the job record to settle from, or the customer contests the close out formally; the record flags the gap and waits, and disputed cases go to you.
Your first run produces: one close out for the first job that cancels after you take the skill on, in about thirty minutes: the job file pulled with the customer, the work, the slot, the total, and the deposit copied from the record, the written cancellation terms copied onto the close out record in full, every material bought for the job listed one line per item and marked as the operator confirms, and the settlement worked in the open, held with a draft closing note and its rebook path for your approval.
Preview, from the skill itself
“# Job Cancellation Handling<br># Job Cancellation Handling Jobs get cancelled.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: b5b40bc6966c… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 33.1 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/fence-build-cancelled-after-materials/WALKTHROUGH.md, examples/roof-valley-repair-cancelled-early/WALKTHROUGH.md, reference/accounting-for-materials-bought.md, reference/reading-the-written-terms.md, reference/releasing-the-slot-and-offering-rebook.md, reference/working-the-settlement-figures.md, templates/cancellation-close-out-record.template.md, templates/cancellation-note-to-customer.template.md
Worked example, from the package
examples/fence-build-cancelled-after-materials/WALKTHROUGH.md
A local operator who builds garden fences. A fence job is booked, materials are bought for it, and the customer then cancels. This is the close out worked from that job record, with the materials list settled line by line.
Year-end stops being a rebuild from scraps: the year's income by source, expenses by class, and vehicle totals are assembled into one packet with every total worked in the open and every gap named as a gap, ready to hand to your accountant.
What changes
Honest fit
For you if: The year's money record is spread across job records, receipts, and a vehicle log, and year end means rebuilding it from scraps for your accountant.
Skip it if: You need tax advice, a filing position, or a missing amount estimated; the packet assembles records only, names its gaps, and sends nothing anywhere.
Your first run produces: one packet for the last finished year, built in one sitting set aside for it: the record index listed month by month across job records, payment records, expense records and receipts, and the vehicle log, with any month holding no records written on the gap list before any totals are built, then income by source, expenses by class, and vehicle totals built in that order, each section total worked in the open as it is finished, and the packet held for your approval.
Preview, from the skill itself
“# Year-End Money Packet<br># Year-End Money Packet At the end of the year the money record for a small operation is usually spread across the whole year and the whole truck: job records in one place, payment notes in another, receipts in envelopes, a vehicle log in a notebook, and a few weeks that live nowhere but memory.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: b67eee30a76c… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 41.2 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/firewood-and-hauling-mixed-records-packet.md, examples/solo-yard-operator-full-year-packet.md, reference/gathering-the-year-records.md, reference/income-by-source-and-expenses-by-class.md, reference/naming-gaps-and-staying-inside-the-records.md, reference/vehicle-totals-from-the-log.md, templates/gap-and-handover-note.template.md, templates/year-end-packet.template.md
Worked example, from the package
examples/firewood-and-hauling-mixed-records-packet.md
An operator who sells firewood deliveries, does hauling and dump runs, and takes tree and brush work when it comes. The records for the year 2025 are uneven: strong in spring and fall, thin in summer, with a second truck used for one week in August and no log kept for it. The packet is built in January 2026 from the records as they stand. It does not smooth the uneven parts away. Income is grouped by line of work, as the job records name it.
Disposal stops being the cost that lands nowhere: every run is planned from the job records, the fee is taken from the site receipt, and each job's share is filed on its own record so the job lines add back to the receipt total.
What changes
Honest fit
For you if: You haul debris from several jobs together, and the disposal fee today lands wherever the day ends rather than on the jobs that filled the load.
Skip it if: You need the site chosen, a ruling on what a site will take, or a price set for the customer; the record holds the receipt fee and its split, and hauling and pricing stay with you.
Your first run produces: one run plan for the next disposal run already coming due, even if it serves a single job, in about twenty minutes: each job the run will serve written from its job record with what that job needs hauled, the site named with anything the operator knows it will not take noted before loading, and, where the run is shared, the split basis written on the plan now, the whole plan held for your approval with nothing loaded from the draft alone.
Preview, from the skill itself
“# Dump and Disposal Runs<br># Dump and Disposal Runs Every cleanup, tear out, and repair leaves something behind: torn off roofing, old fence boards, brush, broken fittings, a shed full of mixed debris.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: ee1fe0a9b7e2… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 33.6 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/shared-run-three-jobs/WALKTHROUGH.md, examples/single-job-roof-tear-off/WALKTHROUGH.md, reference/filing-the-run-against-the-jobs.md, reference/planning-the-run-from-the-jobs.md, reference/recording-the-receipt-and-the-fee.md, reference/splitting-a-shared-run-across-jobs.md, templates/job-disposal-line.template.md, templates/run-plan-and-record.template.md
Worked example, from the package
examples/shared-run-three-jobs/WALKTHROUGH.md
A local service operator finishing three small jobs in the same week: a yard cleanup, a fence tear out, and a shed cleanout. Each job left material behind, none of it a full load on its own. One trailer run served all three on 14 May 2026, planned before loading, with the fee split by load share noted at loading.
A warranty claim starts from the paper, not from memory: proof of purchase, the warranty window checked in the open, dated photos of the failure as found, and the claim drafted in the maker's own terms, tracked until it closes.
What changes
Honest fit
For you if: Gear or materials have failed and the claim will rest on a receipt you may not be able to find and a purchase date held in memory.
Skip it if: You need a ruling that a failure is covered, a promise a claim will succeed, or legal advice on a disputed claim; the paperwork states the terms, the maker decides, and the packet presents the facts in order.
Your first run produces: one claim file for the gear and standing materials the business depends on, in about forty five minutes: the receipts, invoices, and item paperwork pulled into one file, each main item listed with its purchase facts copied from its record and its warranty period quoted from its own paperwork or marked unknown, a tracking log started, and, if a failure is live now, its packet built in the same sitting with the window check worked in the open, photos taken as found, and the draft held for your approval.
Preview, from the skill itself
“# Warranty Claims Packet<br># Warranty Claims Packet Gear fails and materials fail.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: ec522ee9233a… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 40.1 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/circular-saw-motor-failure-claim.md, examples/deck-board-peeling-claim.md, reference/drafting-in-the-makers-terms.md, reference/photos-of-the-failure.md, reference/proof-of-purchase-and-dates.md, reference/tracking-the-claim-to-close.md, templates/claim-packet.template.md, templates/claim-tracking-log.template.md
Worked example, from the package
examples/circular-saw-motor-failure-claim.md
A one person deck and repair operator. The cordless circular saw used on most jobs stops mid cut in September 2026. The claim is built as a packet from the purchase record before the operator speaks to anyone about it.
The letters you send again and again stop drifting: each one lives as a checked template that names its purpose and the facts it requires, filled from the record at send time, so no letter goes out carrying a claim you cannot back that day.
What changes
Honest fit
For you if: You send the same few letters again and again, and each one is rebuilt from the last send with lines nobody has checked since.
Skip it if: You need a one off letter written for a case at hand, legal terms settled, or a letter sent without your read; every filled letter here is a draft held for your approval.
Your first run produces: one short library built from letters you have truly sent, even if it starts with only three or four templates, in about forty minutes: the repeated letters listed from past sends, estimates, and job notes, each with its purpose written in one line, one entry built per letter with every required fact named beside the record that supplies it and every claim listed with its backing, any claim with no backing taken out, and one trial fill worked from a current record to prove the fact list is complete.
Preview, from the skill itself
“# Form Letter Library<br># Form Letter Library Most operators send the same few letters again and again: the estimate follow up, the balance note, the schedule change, the completion letter, the welcome note.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: 804b8bb9d943… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 28.5 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/estimate-follow-up-and-balance-note/WALKTHROUGH.md, examples/schedule-change-and-completion-letters/WALKTHROUGH.md, reference/checking-claims-before-sending.md, reference/filling-facts-from-the-record.md, reference/listing-required-facts.md, reference/naming-each-letters-purpose.md, templates/form-letter-entry.template.md, templates/letter-draft-from-template.template.md
Worked example, from the package
examples/estimate-follow-up-and-balance-note/WALKTHROUGH.md
A small repair operator who fixes porches, steps, and fences. Two letters sit in the library from past sends: Estimate Follow Up and Balance Note. Both were built as entries with their purposes, fact lists, and claims listed before either was used here.
Every change of run or operator starts from a written brief instead of a cold start: the state as checked, the work done, the decisions made, the open items with a holder and a next step, and what to watch, so nothing has to be re-derived or guessed.
What changes
Honest fit
For you if: Work passes from one run or operator to another, and the new run starts by rebuilding what the last one did, decided, and left open.
Skip it if: You need one task handed over as a capsule, the standing record of consequential calls kept, or the work itself done; the brief carries the state at the handover, and the records stay the source each line is checked against.
Your first run produces: one brief at the next handover for one piece of work that changes hands often, in about thirty minutes: an index opened with the scope line written at the top, the first brief written as the current run ends with the state now, the work done, the decisions made, the open items with a holder and a next step for each, and the watch items with the sign for each, any fact no record holds written as not known, and the brief read back by the next run or operator until they can state the state, the open list, and the first step from the brief alone.
Preview, from the skill itself
“# Agent Handover Briefs<br># Agent Handover Briefs Work passes from one run to the next all the time.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: e0131c759f94… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 50.8 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/a-weekend-cover-brief-for-a-repair-shop-inbox/WALKTHROUGH.md, examples/an-overnight-run-handed-to-the-morning-operator/WALKTHROUGH.md, reference/brief-fields.md, reference/keeping-and-correcting-briefs.md, reference/reading-a-brief-on-arrival.md, reference/when-to-write-a-brief.md, templates/handover-brief.template.md, templates/handover-index.template.md
Worked example, from the package
examples/a-weekend-cover-brief-for-a-repair-shop-inbox/WALKTHROUGH.md
A small bicycle repair shop had one weekday operator who ran the inbox and the job board, and a cover person who took both over from Friday evening to Monday morning so the shop could answer customers over the weekend. The cover person was good at the work and new to the shop, and for the first weekends they spent Saturday morning reading back through the week's threads to learn which customers were waiting on what. The operator opened a handover index for the Friday and Monday changes of hands. This is brief 14, written on Friday 9 May, the weekend it covered, and the return brief that closed the loop.
Nothing sensitive sits undecided in silence: every pending approval is kept on one record with its age and its owner, and each item ends decided, escalated, or expired on a rule you set, with the queue swept on a schedule.
What changes
Honest fit
For you if: Sensitive work waits on a yes or a no, and the waits live in heads and inboxes with no age, no owner, and no end.
Skip it if: You need the decision itself made, the held message sent, or owners chased outside the sweep; the queue manages the wait, and the decision and the action stay with their owners.
Your first run produces: one queue built from the queue you have now, in about thirty minutes: the queue opened with its scope line at the top, every place that work currently waits walked, one entry opened per item with what is asked, who asked, who owns it, the date it arrived, its age today, and its expiry rule, and the oldest entry read as a stranger would until what is waiting, who can end the wait, and what happens if they do not can all be said from the entry alone.
Preview, from the skill itself
“# Approval Queue Management<br># Approval Queue Management Work waits for a yes or a no all through the week.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: 6187fec23b30… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 41.2 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/a-weekly-sweep-that-cleared-a-backlog/WALKTHROUGH.md, examples/an-item-that-expired-on-its-rule/WALKTHROUGH.md, reference/aging-and-expiry-rules.md, reference/escalation-routes.md, reference/queue-item-fields.md, reference/reviewing-and-clearing-the-queue.md, templates/approval-queue.template.md, templates/queue-item.template.md
Worked example, from the package
examples/a-weekly-sweep-that-cleared-a-backlog/WALKTHROUGH.md
A two person bookkeeping practice used an agent to draft changes to client records: bank details, billing addresses, and contact names. Every change was held for one of the two partners before it was applied. The holds lived in a shared task list with no dates on them. By early September the list held more than forty waiting changes, some of them months old, and neither partner could say which ones were still wanted. On 8 September they opened an approval queue, with one partner as keeper, covering held record changes only. This is how the backlog was entered, aged, and cleared over four sweeps.
A reviewer who cannot see your live system still gets the full story: your decision and proof trails exported into one dated packet, checked entry by entry before it is shared, with every gap named instead of smoothed over.
What changes
Honest fit
For you if: Someone outside the day to day work needs to check what your system did, and today that means live access and answers from memory.
Skip it if: You need the calls judged, a thin trail repaired, or a dispute settled; the packet carries the trail with its gaps named, and judging it belongs to the reviewer.
Your first run produces: one small packet first, in about fifty minutes: one system, one short period of two to four weeks, and one named reader, the scope line written, the sources listed before anything is pulled and copied as they stand with the pull time noted as the freeze point, the packet built and its cover counts rebuilt from the pages, three entries picked at random and rebuilt from the packet alone, and a gaps page written naming every entry that could not be rebuilt.
Preview, from the skill itself
“# Audit Trail Exports<br># Audit Trail Exports Sooner or later someone outside the day to day work needs to check what a system did.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: 973190cb7b21… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 50.3 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/a-handover-packet-when-the-live-system-was-gone/WALKTHROUGH.md, examples/exporting-a-month-of-approvals-for-an-outside-review/WALKTHROUGH.md, reference/dating-freezing-and-keeping-exports.md, reference/naming-gaps-and-missing-proof.md, reference/packet-contents-and-order.md, reference/proof-checks-before-you-share.md, templates/export-packet.template.md, templates/export-request-and-receipt.template.md
Worked example, from the package
examples/a-handover-packet-when-the-live-system-was-gone/WALKTHROUGH.md
A fictional holiday letting business ran a bookings helper that could change a booking date inside set limits, agree a reply to a guest where the reply followed a standing line, and offer a refund where a stay had fallen short. The helper's trail sat inside a hosted tool the business paid for by the month. In October the owner decided to bring the work in house. The hosted tool would be closed at the end of November, and with it the only copy of the trail, unless a packet was built first. The reader was the owner herself, in her other hat: the person who would answer for the helper's past calls once the tool was gone. This is how the packet was built from what could still be reached, and how its gaps were carried rather than smoothed over.
A release stops depending on memory: it runs step by step against its written checklist, each step evidenced and signed off in order, holds named in the open, and one run record kept per release.
What changes
Honest fit
For you if: You put releases out and the proof of what ran lives in a chat thread and in the memory of whoever ran it.
Skip it if: You need the pre-ship build check, a ruling that a release is ready, or the release published for you; the readiness call is yours, made before the run, and each step still goes through its normal route.
Your first run produces: your next release run start to finish against its written checklist, in one sitting: the checklist walked once before anything starts, each step with its done when line and the evidence it will be signed off on, the run record opened with the product, version, date, run lead, and checklist version, each step done in order with its evidence written and signed off with a name and a time before the next starts, and the record closed with a line stating what went out, when, and which steps, if any, stand as not done.
Preview, from the skill itself
“# Release Checklist Runner<br># Release Checklist Runner A release is a chain of small steps done in an order that matters.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: c673ad41712d… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 39.1 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/a-release-held-at-the-announce-step/WALKTHROUGH.md, examples/a-small-app-release-run-in-order/WALKTHROUGH.md, reference/checklist-shape-and-step-evidence.md, reference/closing-and-keeping-the-run-record.md, reference/holds-skips-and-failed-steps.md, reference/running-steps-in-order.md, templates/release-checklist.template.md, templates/release-run-record.template.md
Worked example, from the package
examples/a-release-held-at-the-announce-step/WALKTHROUGH.md
A small team makes an offline maps app for walkers, put out from its own release page. The team had used a written release checklist for its last three releases, held at version 4 with 12 steps. This run, for version 1.8.1, is the first one the checklist held in the open, and it shows what a hold looks like on a run record: what stopped, what did not start, and how the run went on the next day.
Nobody works from the wrong copy again: every living document has one current version named in a register, superseded copies marked and archived, and one change log line for every version.
What changes
Honest fit
For you if: The same living document exists in several copies that disagree, and nobody can say quickly which one stands.
Skip it if: You need the document's content written or improved, a change agreed, or a lost file recovered; the register records the calls once they are made, and the words stay your work.
Your first run produces: one living document brought under control, in about forty minutes: every copy you can find gathered and laid side by side, the text in use today settled from the records, the settled text given the next clean version name with the control header on its face and its first change log line written, each older copy marked as superseded and moved to the archive, and one older copy checked against the register until you can tell from the copy and the register alone whether it stands and what replaced it.
Preview, from the skill itself
“# Document Version Control<br># Document Version Control Some documents are finished once and never touched again.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: f4eb7d3a00e8… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 44.1 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/client-intake-pack-brought-under-control/WALKTHROUGH.md, examples/price-list-corrected-without-losing-history/WALKTHROUGH.md, reference/choosing-what-needs-control.md, reference/marking-and-archiving-superseded-copies.md, reference/version-names-and-numbers.md, reference/writing-the-change-log.md, templates/change-log-entry.template.md, templates/version-register.template.md
Worked example, from the package
examples/client-intake-pack-brought-under-control/WALKTHROUGH.md
A small house cleaning business with two teams used one living document, a four page client intake pack: service terms, a key handling page, a products page, and a booking form. New clients got the pack by email, a printed stack sat in the office, and each team lead carried a copy. In April the two team leads quoted different terms on the same day, one from the printed stack and one from the file in the shared folder, and the owner could not say which copy stood. The owner opened a version register on 5 May, as keeper, with the intake pack as its first and only entry. This is how that first run went.
Meetings start from the record instead of from memory: the agenda is built before the meeting from the open items and decisions actually owed, each item with one owner and the decision or input it needs.
What changes
Honest fit
For you if: Your meetings start from recall, carry the same ownerless items week to week, and end with the owed list unchanged.
Skip it if: You need notes taken in the room, minutes written, or the meeting chaired; this skill builds the before record, and what happens in the meeting belongs to the meeting record.
Your first run produces: the next agenda for one recurring meeting whose hour often drifts, built from the record in about thirty minutes: the open items gathered from the source records before any are judged, the items needing no decision dropped into a reading list and ownerless items held on the unowned line, each item that stays written up with its owner, the decision or input needed, its source, and its minutes, and the agenda ordered and timed so the minutes fit the meeting, with what does not fit moved to the deferred list.
Preview, from the skill itself
“# Meeting Agenda Builder<br># Meeting Agenda Builder Most meetings start from memory.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: a946f0afca31… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 38.7 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/launch-readiness-meeting-with-decisions-owed/WALKTHROUGH.md, examples/weekly-ops-review-built-from-open-items/WALKTHROUGH.md, reference/agenda-item-fields.md, reference/gathering-open-items-from-the-record.md, reference/timing-and-ordering-the-agenda.md, reference/what-belongs-on-an-agenda.md, templates/agenda-item.template.md, templates/meeting-agenda.template.md
Worked example, from the package
examples/launch-readiness-meeting-with-decisions-owed/WALKTHROUGH.md
Fieldnote is a fictional phone app for site workers to log jobs with photos, built by a fictional four person team. Through the spring the team held a Friday readiness meeting ahead of a planned launch. The early meetings were walk rounds of the build: each person showed what they had finished, the hour passed, and three calls the launch turned on, the launch date, the price page wording, and who would cover support in the first weeks, were still unmade at the start of May. The team lead rebuilt the meeting using this skill. This is the build for Friday 9 May.
An inbox worked in full passes stops being a pile: every message lands in one of four buckets, act now, file, waiting, or discard, the closing counts add back to the start, and waiting items carry an owner and a follow-up date, so nothing important is quietly dropped.
What changes
Honest fit
For you if: Your inbox is a pile worked from the top, and nobody can say which senders are still waiting on an answer.
Skip it if: You need replies sent, messages filed or moved for you, or reply promises set with clients; this skill only sorts, logs, and drafts, and acting stays with you.
Your first run produces: the four buckets set up and the first full pass run in the same sitting, in about forty five minutes: a place made for each bucket in the tool you already use, the pass log and waiting list set up from the templates, the inbox counted with the date and time at the top of the log, each message worked in the order it sits and put in exactly one bucket, a short draft held for each act now item that needs a reply, each waiting item listed with its owner and follow up date, and each discard item listed with its reason.
Preview, from the skill itself
“# Inbox Zero Triage<br># Inbox Zero Triage An inbox that never gets fully sorted stops being a work list and becomes a pile.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: b7dfd1541687… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 29.7 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/contractor-inbox-monday/WALKTHROUGH.md, examples/freelance-desk-friday/WALKTHROUGH.md, reference/defining-the-four-buckets.md, reference/handling-waiting-and-follow-up.md, reference/keeping-nothing-dropped.md, reference/running-a-triage-pass.md, templates/triage-pass-log.template.md, templates/waiting-list.template.md
Worked example, from the package
examples/contractor-inbox-monday/WALKTHROUGH.md
A small roofing and repairs operator whose work inbox gathers messages over the weekend. The pass runs on Monday morning, 11 May 2026, starting at 8:10 in the morning. The operator runs a pass each Monday and Thursday. Names and messages below are sample material for this example; another operator's log holds their own senders and their own counts.
Recurring work stops failing quietly: every job sits on one schedule record with a named owner, a cadence with a due window, and proof of its last run, and any due job with no proof stands marked missed in plain sight.
What changes
Honest fit
For you if: Recurring jobs live scattered across calendars, threads, and heads, and a missed run surfaces weeks later when the gap starts to cost.
Skip it if: You need the jobs run, reminders sent, or one off tasks scheduled; the owners do the work, and this skill keeps the schedule and the proof trail those runs rest on.
Your first run produces: the first schedule record built and the first proof notes filed in the same sitting, in about forty five minutes: the jobs that truly repeat named from the task lists, calendars, and run notes, each job written with what a finished run leaves behind, its owner, its cadence, and its due window, each last run settled from a dated proof note on file or written as NO PROOF ON FILE where none can be found, and the first review date written at the top.
Preview, from the skill itself
“# Recurring Task Scheduler<br># Recurring Task Scheduler Recurring work fails quietly.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: f226a3d4f748… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 32.8 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/monthly-close-review/WALKTHROUGH.md, examples/weekly-ops-checks-review/WALKTHROUGH.md, reference/building-the-schedule-record.md, reference/cadences-and-due-windows.md, reference/last-run-proof-and-gaps.md, reference/reviewing-missed-runs.md, templates/run-proof-note.template.md, templates/schedule-record.template.md
Worked example, from the package
examples/monthly-close-review/WALKTHROUGH.md
A small fictional training provider, Alder Creek Training, with five recurring jobs on one schedule record. Monthly close jobs are read at a monthly review, and the weekly and set day jobs among them are read at the same sitting for the span the review covers. This is the review run on Friday 5 June 2026, covering the runs due from Friday 29 May 2026 to Wednesday 3 June 2026, which is the span the May close falls in.
Your quotes stop resting on memory: your own sell prices for services sit on one written list reviewed on a set cadence, with every change dated and carrying its reason, so each new quote draws from the one current price.
What changes
Honest fit
For you if: Your quotes rest on prices held in memory, copied from quote to quote, with no date and no reason behind any of them.
Skip it if: You need prices set for you, buy prices or supplier costs tracked, or quotes written and altered; every price is your call, and this skill keeps the written record of the prices you settle.
Your first run produces: the first list built and the first review held in the same sitting, in about forty minutes: the services you sell again and again named from past quotes and job records, each written with its unit, a short scope note, and the price the records show was last charged, each line then settled at the review as holds, changes, added, or no longer offered, every change logged with its reason in your own words, and the review cadence and next review date written at the top.
Preview, from the skill itself
“# Price List Maintenance<br># Price List Maintenance A quote is only as steady as the price list behind it, and a price list held in memory is not a list at all.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: edc593298c41… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 33.3 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/midyear-update-hauling-and-repairs/WALKTHROUGH.md, examples/quarterly-review-yard-care/WALKTHROUGH.md, reference/building-the-price-list.md, reference/recording-changes-with-reasons.md, reference/setting-the-review-cadence.md, reference/using-the-list-in-quotes.md, templates/price-change-log.template.md, templates/price-list.template.md
Worked example, from the package
examples/midyear-update-hauling-and-repairs/WALKTHROUGH.md
A fictional local service operator who does hauling runs and small repair work. Four services sit on the price list from past quotes and job records, and the list is reviewed twice a year. This is the review held on 7 July 2026, against the list as it stood from the review on 6 January 2026. At this review one service is added and one is marked as no longer offered.
A customer relationship ends as a finished record instead of an open thread: the final account settled with the arithmetic shown, records handed over on a dated list, standing arrangements closed with end dates, and a plain closing note drafted for your approval.
What changes
Honest fit
For you if: Customer relationships end by drift, with the final account, the handover, and the standing arrangements left wherever they were.
Skip it if: You need legal or tax advice, payments chased or taken, or a formal settlement document; the closing note is a plain courtesy record of an ending, and settling the money stays with you.
Your first run produces: one customer whose work has ended closed whole, in about thirty minutes: the customer file and job records opened and the final balance worked in the open on the checklist, any line no record settles carried as an open line with the reason written down, the handover list built with what the customer gets back and what you keep and why, every standing arrangement listed with its end date or marked open with its next step, and a plain closing note drafted from the template and held for your approval.
Preview, from the skill itself
“# Customer Offboarding Pack<br># Customer Offboarding Pack Work ends in all kinds of ways.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: b0f042e1f2da… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 31.0 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/landscaping-season-close/WALKTHROUGH.md, examples/regular-cleaning-close/WALKTHROUGH.md, reference/closing-access-and-standing-arrangements.md, reference/handing-over-records-and-deliverables.md, reference/settling-the-final-account.md, reference/writing-the-closing-note.md, templates/closing-note.template.md, templates/offboarding-checklist.template.md
Worked example, from the package
examples/landscaping-season-close/WALKTHROUGH.md
A small landscaping operator, Creekside Lawn and Garden, closing out a household customer, Mr. Delaney, after one season of mowing and garden care. The family is moving away at the end of October 2026. The last visit was 9 October 2026, and the ending was worked and filed on 14 October 2026.
Year-end stops being a search: tax documents are filed into one indexed folder as they arrive all year, a monthly sweep names anything missing while it can still be found, and the handover to your accountant is the folder itself. It organizes documents only and gives no tax advice.
What changes
Honest fit
For you if: Tax documents arrive all year and land wherever they land, so year end becomes a search and your accountant gets whatever the search found.
Skip it if: You need tax advice, a return prepared or filed, or totals worked out for the year; it organizes documents only and gives no tax advice, and the handover to your accountant is your own act.
Your first run produces: the folder and the index set up and everything that has arrived so far this year gathered in the same sitting, in about forty five minutes: one folder made with its class list written at the front, the index started with the folder's home, the class list, and the sweep schedule at its head, each document in hand filed under its class with one index line written from the document itself, the expected list written as open items from last year's sources, and the first monthly sweep date set about a month out.
Preview, from the skill itself
“# Tax Document Gathering<br># Tax Document Gathering Tax documents do not arrive at year end.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: 917ae9d0918f… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 45.2 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/full-year-lawn-and-cleanout/WALKTHROUGH.md, examples/midyear-shoebox-start/WALKTHROUGH.md, reference/document-classes-and-labels.md, reference/filing-on-arrival.md, reference/folder-layout-and-the-index.md, reference/monthly-sweep-and-missing-documents.md, templates/document-index.template.md, templates/handover-note.template.md
Worked example, from the package
examples/full-year-lawn-and-cleanout/WALKTHROUGH.md
A one person lawn care and property cleanout operator, working under the name Mercer Lawn and Cleanout. He does the work alone, keeps a vehicle log in a notebook in the truck, and uses an accountant at year end. This is tax year 2025, gathered from a first run on 6 January 2025 to the handover on 30 January 2026.
Praise you already earned stops getting lost or misused: reviews, messages, and comments are gathered into one library with the permission state recorded on each item, so a testimonial goes public only when its record covers that use.
What changes
Honest fit
For you if: Kind words about your work sit scattered across reviews, messages, and comments, with no record of what each person agreed to.
Skip it if: You need praise written, reviews asked for, or anything posted; it collects praise that already exists and nothing else, and public use stays with you after a check.
Your first run produces: the first library built from praise that already exists, in about thirty minutes: the reviews, messages, and comments you supply worked through in the order you supply them, one entry made per item with its exact words, who said it, where it was left, the date, and what work it was about, and a permission state set for each entry from the record only, with items the record does not cover set to held and their reasons written in plain words.
Preview, from the skill itself
“# Social Proof Collector<br># Social Proof Collector Praise for good work already exists.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: c93f4fdce327… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 33.8 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/creator-course-feedback/WALKTHROUGH.md, examples/local-workshop-client-notes/WALKTHROUGH.md, reference/checking-coverage-before-any-use.md, reference/finding-praise-that-already-exists.md, reference/keeping-the-library-honest-and-current.md, reference/recording-permission-per-item.md, templates/proof-library-entry.template.md, templates/public-use-check.template.md
Worked example, from the package
examples/creator-course-feedback/WALKTHROUGH.md
A fictional illustration teacher, Mara Ellis, runs a small online drawing course. Over one term, kind words about the course landed in five places. She wants a testimonial section on her course page, built only from praise whose record covers that use.
A large job stops failing one unnamed step at a time: the whole job is broken into ordered steps, each carrying what it needs before it can start, who owns it, and a written test for done, kept as one living plan the work is checked against as it runs.
What changes
Honest fit
For you if: A large job runs in pieces, with the order in one person's head and done meaning different things to different people.
Skip it if: You need the steps themselves done, people assigned, dates moved, or time and cost tracked; the plan holds the outcome, the order, and the test, and the work itself stays with the step owners.
Your first run produces: the first plan for a real job that is already waiting or already running, in about an hour: the job written in one outcome line a stranger could read, one entry per step with its owner, the steps it depends on, and a done criterion a second person could check, the steps ordered from their dependencies alone, and the first check date set before you stop.
Preview, from the skill itself
“# Multi-Step Job Planner<br># Multi-Step Job Planner A large job rarely fails all at once.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: d4880e28e207… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 51.2 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/help-article-portal-move-in-nine-steps/WALKTHROUGH.md, examples/twelve-lesson-training-pack-built-to-a-plan/WALKTHROUGH.md, reference/keeping-the-plan-current.md, reference/ordering-steps-and-dependencies.md, reference/plan-terms-and-parts.md, reference/writing-done-criteria-that-hold.md, templates/job-plan.template.md, templates/step-entry.template.md
Worked example, from the package
examples/help-article-portal-move-in-nine-steps/WALKTHROUGH.md
A four person support team held 214 help articles in an old folder set and was moving the live ones to a new portal with a different section shape. Two earlier tries had started with the move itself and stalled when articles were found mid move that nobody had counted, mapped, or decided to keep. This time the team planned the job first, in nine steps, and checked the work against the plan twice a week. The names and figures here are a worked example, not a record of a real move.
Know what your system does in the first minute after a part fails, before a real failure shows you in front of a client: each failure is scripted in advance, run on copies or in a sandbox, and the observed behavior is recorded against the behavior written down before the run.
What changes
Honest fit
For you if: You cannot say what your system does in the first minute after a part fails, and the first real outage would be the first rehearsal.
Skip it if: You need hostile input testing, live monitoring, or the faults fixed; drills rehearse faults on copies only, name the gaps, and leave attacks, alarms, and fixes to their own work.
Your first run produces: the first drill, run on a copy, in about forty five minutes: the record opened for the part whose failure would hurt most this month, one script of six to ten steps with the expected behavior written before the run, the smallest copy built with no route to the live system, the script run once with observed behavior written beside expected as it happens, and the date set for a second drill against a different part.
Preview, from the skill itself
“# Failure Mode Drills<br># Failure Mode Drills Most agent systems are only ever watched while everything they lean on is working.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: c6f58d60458e… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 52.5 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/bad-input-batch-at-intake/WALKTHROUGH.md, examples/provider-outage-mid-run/WALKTHROUGH.md, reference/failure-modes-and-expected-behavior.md, reference/reading-drill-results.md, reference/running-on-copies-and-in-a-sandbox.md, reference/scripting-a-drill.md, templates/drill-record.template.md, templates/drill-script.template.md
Worked example, from the package
examples/bad-input-batch-at-intake/WALKTHROUGH.md
A bookkeeping helper agent took in a batch of client files each week, read each file, and filed its rows into a records store. Good weeks had made the intake look solid. The operator wanted to know what one bad batch would do: a batch carrying an empty file, a malformed file, and a file that had already been sent once before. Rather than wait for that batch to arrive in real work, she built it herself, from copies, and ran it through a copy of the intake.
Access stops drifting one quiet grant at a time: every grant a person, an agent, or a connected tool holds sits in one standing register and is rechecked on its due date, then confirmed, narrowed, or revoked on the record, on a fixed rhythm instead of after something goes wrong.
What changes
Honest fit
For you if: Grants held by people, agents, and tools have piled up unchecked, and nobody can say which are still used, still needed, or still the right width.
Skip it if: You need a first full inventory of everything an agent can reach, new access handed out, or watching between cycles; this is the recurring rhythm that keeps a known register honest, and changes are made elsewhere.
Your first run produces: the standing register and the first cycle run over it, in about an hour: one line per grant with its holder, what it reaches, its width, its owner, and a first due date, gaps written as not yet known and flagged rather than held back, the cadence chosen for the size of the list, every due grant called confirmed, narrowed, or revoked on the record, and the next cycle date written at the top of the register.
Preview, from the skill itself
“# Access Review Cadence<br># Access Review Cadence A one time look at access goes out of date the week it is finished.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: 521849d7e1d5… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 43.0 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/monthly-cycle-for-a-small-support-team/WALKTHROUGH.md, examples/quarterly-cycle-that-ended-stale-grants/WALKTHROUGH.md, reference/cadence-terms-and-the-register.md, reference/choosing-the-cadence.md, reference/confirm-narrow-or-revoke.md, reference/stale-grants-and-drift-signs.md, templates/grant-review-entry.template.md, templates/review-register.template.md
Worked example, from the package
examples/monthly-cycle-for-a-small-support-team/WALKTHROUGH.md
A four person support team, with one agent handling first replies, kept its access register on a monthly full cycle. The list was short enough to read whole each month, and holders changed often enough that a quarter felt too long to wait. This example follows the first three cycles, and reads the third one in full, in counts.
Every incoming incident is graded against one written severity scale, with the grade and its reasons on record and the response target following from the grade, so a fault that stops buyers paying stops sharing a queue with a spelling slip.
What changes
Honest fit
For you if: Incidents are taken in arrival order, so a loud minor fault outranks a quiet serious one and nobody can say why one waited.
Skip it if: You need incident messages drafted, a postmortem written, or the fault itself fixed; this grades and sets the target only, and the response work stays with its own skills.
Your first run produces: a first severity scale tested against the record, in about an hour: the scale built with four grades and a response target for each that the team can meet on an ordinary day, three or four closed incidents graded on their own entries using only the facts known when each first arrived, the entries read side by side and the scale adjusted once where the order comes out wrong, and the scale review date set before you stop.
Preview, from the skill itself
“# Incident Severity Grading<br># Incident Severity Grading When incidents arrive with no shared scale, attention goes to whoever reported last, whoever reported loudest, or whoever sounds most worried.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: 7272c2ecc266… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 37.8 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/morning-when-the-store-went-down-and-a-typo-waited/WALKTHROUGH.md, examples/slow-editor-and-the-loud-print-fault/WALKTHROUGH.md, reference/grading-by-impact-and-reach.md, reference/regrading-and-borderline-cases.md, reference/response-targets-per-grade.md, reference/the-severity-scale.md, templates/grading-entry.template.md, templates/severity-scale.template.md
Worked example, from the package
examples/morning-when-the-store-went-down-and-a-typo-waited/WALKTHROUGH.md
Harbour Lane is a fictional small studio that sells printable planners through its own store page. On a Tuesday morning two reports arrived eleven minutes apart. One stopped every sale the studio could make. One was a misspelled word. This is how both were graded against the same scale, on the same morning, by the same grader, and why the second one waiting was the scale working, not the report being lost.
Records held past their retention period stop being a habit nobody can explain: each approved hold is written down as one exception naming the records it covers, the reason, the owner, and the date it ends, and is reviewed on the day it lapses.
What changes
Honest fit
For you if: Records are held past their retention period as a habit nobody can explain, with no owner and no end date.
Skip it if: You need retention periods set, changed, or deletion passes run; the schedule does that work, and this register only records stated deviations from it, each with an expiry.
Your first run produces: the first exception register, in about thirty minutes: the schedule where records are most often held past their period read line by line, every current hold found and written down including the informal ones, each hold counted from the store itself and given an owner who states the reason in their own words, holds no owner will stand behind closed and their records put back under the schedule, and each real hold entered and put to the operator for approval as written.
Preview, from the skill itself
“# File Retention Exceptions<br># File Retention Exceptions A retention schedule states how long each record type is kept.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: 42dd2e95a69e… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 45.2 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/project-drafts-held-for-a-client-query/WALKTHROUGH.md, examples/reviewer-notes-renewed-then-closed/WALKTHROUGH.md, reference/approving-and-refusing-exceptions.md, reference/exception-terms-and-states.md, reference/lapse-review-and-closure.md, reference/the-exception-register.md, templates/exception-entry.template.md, templates/lapse-review-log.template.md
Worked example, from the package
examples/project-drafts-held-for-a-client-query/WALKTHROUGH.md
A two person drafting and summary service keeps a retention schedule under Data Retention Schedule. One line on it covers delivery drafts: kept for 60 days from the delivery date, then held for an approved pass. In August a client raised a billing query on one finished project, and the drafts for that project were about to pass their period while the query was still open. This is how the hold was handled as an exception, from the check that found it to the closure that ended it.
A workbook's numbers earn trust from its formulas, not its displayed values: every formula in scope is read range by range for errors, typed values hidden inside calculation chains, and references pointing at nothing, with each fault listed per sheet by cell and class.
What changes
Honest fit
For you if: A workbook's totals are doubted, or about to be handed on, and reading the displayed values cannot settle whether the formulas behind them are sound.
Skip it if: You need the data tidied, a report written from the sheet, or the formulas fixed; the audit names faults for the operator to decide on, and every fix is separate work.
Your first run produces: one sheet audited on a copy, in about half an hour: the workbook copied and the copy named as the audit copy, the audit record opened with the sheet and ranges in scope and the agreed constants named, the cells in scope counted first so the reading has a total to answer to, each range read formula by formula with every fault written as a finding with its cell, formula, class, and effect, and the counts reconciled with the ranges not read named as not read.
Preview, from the skill itself
“# Spreadsheet Formula Audit<br># Spreadsheet Formula Audit A workbook can look right and add up wrong.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: f1a988efc474… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 57.2 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/monthly-budget-with-a-broken-total/WALKTHROUGH.md, examples/stock-count-with-hardcoded-prices/WALKTHROUGH.md, reference/error-and-reference-faults.md, reference/formula-terms-and-scope.md, reference/hardcoded-values-and-chains.md, reference/reading-findings-by-severity.md, templates/audit-record.template.md, templates/sheet-findings-entry.template.md
Worked example, from the package
examples/monthly-budget-with-a-broken-total/WALKTHROUGH.md
A fictional one person workshop kept its monthly budget in a workbook with three sheets: Inputs, Spending, and Summary. The Summary sheet said 5 was left over for the month. The operator doubted it, because the spending on the ground felt heavier than the plan. This is how the workbook was audited, sheet by sheet, and what the audit found. No formula was changed during the audit. The findings went to the operator in the audit record, and fixes were left as separate work.
Every quote you send carries the same current terms: the payment, changes, warranty, and validity blocks live in one checked library, each with its own version, and every new quote is checked against that library before it goes out, so stale wording stops travelling from quote to quote.
What changes
Honest fit
For you if: Your quotes carry terms copied from the last quote, so stale wording travels and a change you made once for one job reads as your standard.
Skip it if: You need legal advice, custom terms written from nothing, or prices and scope checked; the library organizes and checks wording you already hold and settle, and sending stays with you.
Your first run produces: the library built from past quotes, in about forty minutes: the last five quotes the operator sent gathered and worked from only, each kind of terms block lifted out and its versions laid side by side, agreeing wording settled as the starting current version and disagreements listed as gaps for the operator to settle rather than picked by the agent, each settled block given version 1.0 and the date of the run, and the index held for the operator to read before any block becomes current.
Preview, from the skill itself
“# Quote Terms Library<br># Quote Terms Library Every quote carries more than a price.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: 1142c0f077c4… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 35.0 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/kitchen-remodel-terms-caught-before-sending/WALKTHROUGH.md, examples/roof-repair-library-first-build/WALKTHROUGH.md, reference/building-the-library-from-past-quotes.md, reference/checking-a-quote-before-it-goes-out.md, reference/keeping-terms-honest-and-in-scope.md, reference/versioning-blocks-and-retiring-stale-ones.md, templates/library-index-and-check-log.template.md, templates/terms-block.template.md
Worked example, from the package
examples/kitchen-remodel-terms-caught-before-sending/WALKTHROUGH.md
A local service operator who fits and remodels kitchens. The library had stood for two months. A quote for a customer, Dana Whitfield, to remodel a small kitchen was drafted from a saved quote file instead of from the library. This example follows the check that caught two stale blocks on that draft, how each was settled, and the quote that went out carrying the current blocks.
Who had which tool, on which job, and what came back damaged is answerable from one log: your own gear is checked out to a named person for a named job and back in against the same entry, with loss and damage written on that entry.
What changes
Honest fit
For you if: Tools go out with the crew on a word, and when something comes back damaged or does not come back, nobody can say who had it or on which job.
Skip it if: You need rented or borrowed gear tracked, repairs costed, or fault decided; the log states who had a tool, on which job, and the state it came back in, and those calls stay with you.
Your first run produces: the log opened on the next job's tools, in about twenty five minutes: each owned tool or kit listed as its own item, named so it can be told apart from the next one like it, the person and the job written in plain words for each, the condition noted as it stands today, tools that came back before the log started marked plainly as a first note rather than a check out note, and every returned tool checked in on the same entry or listed as still out.
Preview, from the skill itself
“# Tool Checkout Log<br># Tool Checkout Log A crew's own tools move all day.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: f20332d8735c… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 31.7 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/fence-line-clearing-loppers-did-not-come-back/WALKTHROUGH.md, examples/porch-repair-saw-back-with-a-cracked-guard/WALKTHROUGH.md, reference/checking-back-in-on-the-same-entry.md, reference/checking-out-against-a-person-and-a-job.md, reference/reading-the-log-for-what-is-still-out.md, reference/recording-loss-and-damage.md, templates/checkout-slip.template.md, templates/tool-checkout-log.template.md
Worked example, from the package
examples/fence-line-clearing-loppers-did-not-come-back/WALKTHROUGH.md
A local service operator who clears fence lines and brush, with a crew of three. This example follows one clearing job where five items went out and four came back. The fifth was written as not back on its own entry, checked over four days, and set as lost by the operator only once those checks were done and written down.
A contested invoice gets settled from the record instead of argued from memory: each disputed line is checked against the accepted quote and the completion record, the reply is drafted in plain words from what the records show, and the settlement is recorded on the job.
What changes
Honest fit
For you if: A customer contests lines on an invoice and settling it means arguing from memory against theirs.
Skip it if: You need invoices built, undisputed invoices chased, or legal advice; this starts only when a charge is contested, and the settlement itself is your call.
Your first run produces: one contested invoice worked from the records, in about thirty minutes: the most recent disputed invoice taken, its contested lines listed in the customer's own words, each line checked against the accepted quote and the completion record on the check sheet with every figure traced, the answer drafted in plain words from the findings, and the settlement entry held ready to complete once the operator decides.
Preview, from the skill itself
“# Invoice Dispute Handling<br># Invoice Dispute Handling A dispute starts when a customer looks at an invoice and says: that line is not right.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: 6461c851645f… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 37.8 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/gutter-job-elbow-line-held-pipe-line-removed/WALKTHROUGH.md, examples/yard-clearance-third-load-and-an-hour/WALKTHROUGH.md, reference/checking-each-line-against-the-records.md, reference/drafting-the-answer-in-plain-words.md, reference/reading-the-dispute-and-opening-the-record.md, reference/settling-and-recording-the-outcome.md, templates/dispute-check-sheet.template.md, templates/settlement-record.template.md
Worked example, from the package
examples/gutter-job-elbow-line-held-pipe-line-removed/WALKTHROUGH.md
A local service operator who fits and repairs gutters. A customer, Dana Whitfield, had a side gutter run and a downspout replaced. This example follows a dispute over two parts of the invoice: an extra pair of downspout elbows, and a line for an extra downspout pipe length. One part stood on the record. The other did not, and came off.
You stop feeding the sources that never send work: every lead is recorded with where it came from, outcomes are joined back to sources on a set review cadence, and effort goes to the sources your own record supports.
What changes
Honest fit
For you if: You feed every source alike because no record says which ones actually send work that wins.
Skip it if: You need leads generated, listings run, or a best source named in general; the log reads one operator's own record for one named period and says no more than that record holds.
Your first run produces: the log opened and the recent record entered honestly, in about thirty minutes: the fixed source list written at the head of the log, the leads already held entered one at a time from message threads, the job book, or quote records, each with one source or an honest unknown where no record names one, ended leads closed with their outcomes and live ones left open, and a first tally worked for the period with the parts checked against the whole.
Preview, from the skill itself
“# Lead Source Tracking<br># Lead Source Tracking Category: Local Services / Sales”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: 32c12f100138… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 40.0 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/marketplace-heavy-month-for-a-gutter-operator/WALKTHROUGH.md, examples/spring-review-for-a-fence-operator/WALKTHROUGH.md, reference/acting-on-the-review-without-overreading-it.md, reference/recording-outcomes-and-joining-them-back.md, reference/recording-the-source-at-intake.md, reference/running-the-source-review.md, templates/lead-source-log.template.md, templates/source-review.template.md
Worked example, from the package
examples/marketplace-heavy-month-for-a-gutter-operator/WALKTHROUGH.md
A local operator who cleans, repairs, and replaces house gutters. The operator had a feeling that the marketplace listing was the busy channel and the one worth the most evening time. This example follows one monthly review that set that feeling beside the joined record for the month, counted from the month's own lead list.
Each version ships with notes a user can scan in seconds: the change record is rewritten in plain words about what changed for the user, grouped as new, improved, and fixed, with anything the record does not support left out, and every draft held for your approval.
What changes
Honest fit
For you if: Each version ships with notes written from memory at the last minute, so quiet fixes vanish and changes that slipped get announced early.
Skip it if: You need the change record kept, a launch announcement written, or the notes published; this drafts the per version notes from the record and holds them for your approval.
Your first run produces: one version's notes drafted from its record, in about thirty minutes: the change record for the next version to ship read from its first entry to its last, the entries a user can feel set out and the rest written down as left out candidates, the draft written with an opening line naming the version and its kind, then the new, improved, and fixed lines in user words, and the check log filled in as you go, one line for each line in the notes and one for each entry left out.
Preview, from the skill itself
“# Release Notes Drafts<br># Release Notes Drafts Release notes are the short page of notes that ships with, or sits beside, one version of a product.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: 6837a0fe86b5… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 33.2 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/a-fix-only-release-with-nothing-new/WALKTHROUGH.md, examples/a-mixed-release-with-new-improved-and-fixed/WALKTHROUGH.md, reference/grouping-and-ordering-the-notes.md, reference/holding-the-draft-for-approval.md, reference/reading-the-change-record.md, reference/turning-changes-into-user-words.md, templates/release-notes-check-log.template.md, templates/release-notes-draft.template.md
Worked example, from the package
examples/a-fix-only-release-with-nothing-new/WALKTHROUGH.md
The same invented studio from the first example ships Fernhollow Tasks 3.2.1, a small version that follows 3.2 by two weeks. It holds no new feature. This example shows the draft for a release like that: the Fixed group standing alone, the opening line saying so plainly, and no line stretched to fill a New or Improved heading.
The questions your audience actually asks stop getting lost: each one is kept in the asker's own words with where and when it was asked, grouped and counted from the bank itself, then turned into FAQ drafts first, Q and A content second, and product fixes where the same question keeps exposing the same gap.
What changes
Honest fit
For you if: Your audience asks the same questions again and again across comments, messages, and reviews, and each one is answered once and lost.
Skip it if: You need questions scraped from places you do not ordinarily see, replies sent, or results promised; the bank holds questions and counts from your own channels, and every use of it is a draft held for your approval.
Your first run produces: the bank opened on the last two weeks of questions, in about thirty minutes: the questions the operator can already see taken from recent comments, messages, reviews, and counter notes, each entered in the asker's own words with where and when it was asked and nothing entered from memory, the entries grouped by theme and each theme counted from the entries themselves, and two or three FAQ lines drafted for the theme with the highest count, with answers read from the operator's own facts and unsettled points left as named gaps.
Preview, from the skill itself
“# Audience Question Bank<br># Audience Question Bank The questions an audience asks are already market research, written in the audience's own words.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: 60adaca8a91e… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 39.1 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/bakery-counter-questions-that-became-an-faq/WALKTHROUGH.md, examples/bike-shop-questions-that-fixed-a-gap/WALKTHROUGH.md, reference/capturing-questions-in-their-own-words.md, reference/grouping-and-counting-questions.md, reference/turning-questions-into-content-and-fixes.md, reference/turning-the-bank-into-an-faq-page.md, templates/faq-and-content-draft.template.md, templates/question-bank.template.md
Worked example, from the package
examples/bakery-counter-questions-that-became-an-faq/WALKTHROUGH.md
A small bakery with a shop counter and a Saturday market stall, selling bread, cakes, and bakes to order. Over four weeks the operator kept a question bank from the channels she sees in her work: the counter, the stall, comments on the bakery page, and messages sent to the bakery. This example follows that bank from its entries, through its counts, to an FAQ draft held for her approval.
A period of agent work gets a written account instead of a fresh guess: what ran, what the records show went well or badly, and lessons kept as dated notes with the change each one produced.
What changes
Honest fit
For you if: Your agents do steady work week after week, and at the end of a period nobody can say from the records what went well, what kept going wrong, or what a fix changed.
Skip it if: You need a single incident worked in depth or a single decision settled; a postmortem takes one event and a decision log holds one call, and this review reads across a whole period instead.
Your first run produces: the most recent full week of agent work reviewed on its own, in about an hour: the scope line written at the top naming the agents covered, the records that count as sources, and the period dates, what ran in that week gathered and counted as total runs, runs that ended well, runs that ended badly, and runs with no recorded outcome, the poor outcomes grouped and read together, one lesson note drafted where the same cause sits behind three or more of them, one short note written for the strongest thing that went well, and each change agreed with its owner and start date, with the next review date written at the top.
Preview, from the skill itself
“# Agent Retrospective Notes<br># Agent Retrospective Notes Agent work adds up quietly.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: 6bd55bc70e1b… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 43.6 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/a-busy-month-that-changed-how-reminders-were-sent/WALKTHROUGH.md, examples/a-quiet-quarter-that-still-changed-two-rules/WALKTHROUGH.md, reference/lesson-notes-fields.md, reference/reading-a-period-for-lessons.md, reference/retrospective-scope-and-sources.md, reference/tracking-changes-to-the-next-period.md, templates/lesson-note.template.md, templates/retrospective.template.md
Worked example, from the package
examples/a-busy-month-that-changed-how-reminders-were-sent/WALKTHROUGH.md
A small home repair office ran two agents: a dispatch agent that booked jobs, held the job records, and sent appointment reminders the day before each job, and an inbox agent that drafted replies to customers. The owner ran both and kept their records in one place. Jobs were getting done, but reminders kept drawing replies that said the day or the time was wrong, and each reply was handled on its own as a small mix up. At the end of March the owner ran the first monthly retrospective, covering 1 March to 31 March, to read the month as a whole. This is how that review was built and what it changed.
Every prompt change stops being a mystery: each version is kept whole in one registry with what changed, why, and which runs used it, and a named earlier version can be brought back whole when a change makes things worse.
What changes
Honest fit
For you if: Your agents run under prompts that get edited in place, so a week later nobody can say what changed, when, or which wording shaped a questioned run.
Skip it if: You need prompt wording written or judged, or configuration versioned; this registry holds prompt versions, change notes, and run stamps, and moving the live mark still waits for your approval.
Your first run produces: the registry started for one agent and its most shaping prompt set, in about forty minutes: the scope line written naming the agent and the prompts it covers, the wording as it stands today entered as the first live version with its full text copied exactly, the honest reason written on the entry with no earlier history rebuilt, the run stamp set so each run record carries the prompt name and version label, and the next run checked to show the stamp in place.
Preview, from the skill itself
“# Prompt Version Registry<br># Prompt Version Registry An agent behaves the way its prompts tell it to behave.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: 88056c1dda87… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 52.2 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/bringing-an-old-prompt-back-after-a-bad-change/WALKTHROUGH.md, examples/tracing-which-prompt-version-ran-a-job/WALKTHROUGH.md, reference/linking-runs-to-versions.md, reference/naming-and-numbering-versions.md, reference/registry-entry-fields.md, reference/rollback-and-retirement.md, templates/prompt-registry.template.md, templates/prompt-version-entry.template.md
Worked example, from the package
examples/bringing-an-old-prompt-back-after-a-bad-change/WALKTHROUGH.md
A small tiling business ran a quoting helper that read the fitter job notes for a finished visit and drafted the customer quote: the work done, the materials used, and the price lines from the price record. The helper ran under one prompt set, the quote draft prompt, held in a registry the owner kept. By the start of June the registry held six versions, V1 to V6, with V6 live. This is the change that went wrong in June, the rollback to a named earlier version, and what the runs showed afterwards.
Each tool an agent can touch is scoped to the least access its stated job needs, worked out per task type and written down on one sheet, with every wider exception approved and dated.
What changes
Honest fit
For you if: Your agents use tools that were connected in a hurry and still hold the wide reach their setup pages offered first.
Skip it if: You need standing grants audited or re-reviewed on a rhythm; this skill writes the stated scope those checks later read against, before a grant is made or a pairing is added.
Your first run produces: one agent's most used tools scoped on one sheet, in about an hour: the three or four task types the agent does most listed in plain work words, the job of each tool inside each task type written in one sentence, each pairing set with the least scope tests and written as its own scope line with the date set, any pairing that must stand wider written as an exception with its approval and end date, and the settings read back from their settings pages to match the sheet.
Preview, from the skill itself
“# Tool Permission Scoping<br># Tool Permission Scoping A tool is usually connected in a hurry.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: 28eb75083623… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 47.6 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/adding-a-payment-tool-with-a-tight-scope/WALKTHROUGH.md, examples/scoping-a-new-inbox-and-files-helper/WALKTHROUGH.md, reference/exceptions-and-dated-approvals.md, reference/least-scope-tests.md, reference/task-types-and-tool-jobs.md, reference/writing-the-scope-sheet.md, templates/exception-request.template.md, templates/tool-scope-sheet.template.md
Worked example, from the package
examples/adding-a-payment-tool-with-a-tight-scope/WALKTHROUGH.md
A small house cleaning business ran an invoicing helper that raised invoices from the job records and followed up the late ones. In March the owner added a payment tool, so customers could pay an invoice from the invoice itself instead of by bank transfer days later. The tool arrived with a wide first offer: it could raise a charge for any amount, read the payment state of every charge on the account, and pay money out of the account as well as take it in. None of that width was scoped, and the work had not started. This is how the tool was scoped before its first charge, the one exception its first month needed, and the one wider ask that was refused and kept on file.
A release reaches everyone in stages instead of all at once: a small first audience, one fixed check list run at each stage, and a written widen, hold, or roll back call before it moves further.
What changes
Honest fit
For you if: Your releases go to everyone at once, so a new fault gets its largest possible audience on its first day.
Skip it if: Your release cannot be served to part of its audience at a time; this skill only works where a release can in fact be served in parts, and it does not serve any version itself.
Your first run produces: the stage plan opened for the next release, in about forty-five minutes of setup: the stages written with their shares and watch periods, the new version and the prior version named as the roll back target, the check list written for the product with the source and break line for each check, the roll back route tested before the first stage opens, and the first stage opened to the first audience only, with one short check round to run at the end of each stage.
Preview, from the skill itself
“# Canary Release Checks<br># Canary Release Checks A release that goes to everyone at once gives a new fault its largest possible audience on its first day.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: 48e658dbe85f… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 48.8 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/a-clean-widening-that-reached-everyone/WALKTHROUGH.md, examples/a-hold-then-rollback-at-the-second-stage/WALKTHROUGH.md, reference/running-a-rollback-and-after.md, reference/stage-plan-fields.md, reference/the-fixed-check-list.md, reference/widen-hold-or-rollback-calls.md, templates/canary-stage-plan.template.md, templates/stage-check-record.template.md
Worked example, from the package
examples/a-clean-widening-that-reached-everyone/WALKTHROUGH.md
A small team runs a fictional field notes app for garden crews, called Field Notes in this example. Crews note jobs on phones and the notes sync when a signal returns. Version 4.2 added a faster sync and moved the photo add button. Version 4.1 was the prior version in wide use. The release lead, acting as keeper, opened a canary stage plan on 6 April, before version 4.2 was served to anyone outside the team. The operator was the studio owner. This is how the widening ran, stage by stage, to everyone.
Tables trapped in PDFs become spreadsheets you can check: rows and totals lifted page by page and reconciled against the totals printed in the source, with every unreadable cell flagged in place rather than guessed.
What changes
Honest fit
For you if: You hold tables inside PDFs that you need in a spreadsheet you can sort, sum, and check, and hand copying keeps letting small errors through.
Skip it if: You need the figures in the source judged or acted on; reconciliation proves the copy matches the source, not that the source was true, and no payment or correction follows from this skill.
Your first run produces: one short PDF table extracted and proven end to end, in about forty-five minutes: the run sheet opened with the source file named, the page map filled in page by page with headers, data, subtotals, and footers marked, the table extracted with a source page column on each row, the first doubtful cell flagged in place as it is met, and the reconciliation report worked, with data rows counted on both sides and each printed total summed and set against the source.
Preview, from the skill itself
“# PDF Table Extraction<br># PDF Table Extraction A table inside a PDF looks finished.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: 31bff007ad2e… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 46.9 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/a-rotated-statement-table-with-flagged-cells/WALKTHROUGH.md, examples/a-three-page-stock-list-reconciled-to-its-totals/WALKTHROUGH.md, reference/reading-cells-and-flagging-unreadable-ones.md, reference/reconciling-rows-and-totals.md, reference/spreadsheet-output-and-column-types.md, reference/table-discovery-and-page-map.md, templates/extraction-run-sheet.template.md, templates/reconciliation-report.template.md
Worked example, from the package
examples/a-rotated-statement-table-with-flagged-cells/WALKTHROUGH.md
Bramble Lane Bakery is a fictional bakery that buys flour, yeast, and sugar on account from a fictional supplier, Colebrook Milling Co. The supplier sends a monthly account statement as a five page PDF. The bakery owner, the operator for this work, wanted one month in a spreadsheet so the statement lines could be matched against the delivery notes in date order. The statement table was the hard kind: it ran across two pages, and the second page was printed sideways.
A new project starts with names that hold up: one written naming convention for files, versions, and deliverables, with checker rules that test each new name before it is saved.
What changes
Honest fit
For you if: You are starting a project or a team and want its files named well from the first week, before the pile forms.
Skip it if: You need an existing library renamed or tidied; that is a filing run under its own manifest and approval, and this skill designs the convention for something new.
Your first run produces: the convention designed for one new project before it fills up, in about an hour: the kinds of thing the first month will name listed, the parts of a name settled in their order with the date form and numbering rule, versions and status marks settled, the checker built one rule per line, and the whole draft tested against at least ten names that should pass and ten that should fail before the first real files are named under it.
Preview, from the skill itself
“# Naming Convention Setup<br># Naming Convention Setup A new project starts clean for about a week.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: fab2535d62f5… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 53.2 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/designing-a-convention-for-a-new-research-study/WALKTHROUGH.md, examples/starting-a-shared-drive-for-a-new-team/WALKTHROUGH.md, reference/checker-rules.md, reference/parts-of-a-good-name.md, reference/rolling-out-and-revising.md, reference/versioning-and-status-marks.md, templates/convention-sheet.template.md, templates/name-checker.template.md
Worked example, from the package
examples/designing-a-convention-for-a-new-research-study/WALKTHROUGH.md
A four person field research team was about to start a three month study of water quality along a creek, for a county client. The study had a name, a start date, and an empty shared folder. It had no files yet, which was the whole reason the owner ran this skill first: the convention could be designed before the first field note set the pattern by accident.
Every sponsor slot in your newsletter is booked against a named issue in one dated inventory, with copy and link due dates tracked per booking and delivery confirmed from the issue that actually went out, so a paid slot is never double booked or run unproven.
What changes
Honest fit
For you if: You sell sponsor slots in a newsletter and the bookings live across messages, rate notes, and memory, so a slot can be double booked or run unproven.
Skip it if: You need sponsors found, slots pitched or priced, or sponsor copy written; this skill records the bookings and prices you settle and confirms delivery from the issue that went out.
Your first run produces: the slot inventory built for the next four issues and the ledger opened in the same sitting, in about an hour: the slots each issue carries named from the rate notes and recent issues with the price held for each, the publication's disclosure rule written into the ledger head in its own words or flagged as unsettled, four issues laid out by number and date with every slot shown as booked, held, or open, and any booking the records already hold entered whole with its copy and link due date.
Preview, from the skill itself
“# Newsletter Sponsor Slots<br># Newsletter Sponsor Slots A sponsor slot is a promise tied to one issue on one date.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: 6f413a259cb9… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 33.5 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/autumn-launch-four-issues/WALKTHROUGH.md, examples/spring-season-six-issues/WALKTHROUGH.md, reference/booking-slots-against-issues.md, reference/building-the-slot-inventory.md, reference/disclosure-and-delivery-confirmation.md, reference/tracking-copy-links-and-due-dates.md, templates/issue-delivery-check.template.md, templates/slot-inventory-ledger.template.md
Worked example, from the package
examples/autumn-launch-four-issues/WALKTHROUGH.md
A fictional letter for people who run small harbour side shops and stalls, going out every two weeks. Each issue carries two sponsor slots. This walkthrough covers the first four issues after the slots were first offered: Issue 12 on 3 October 2026 to Issue 15 on 14 November 2026, one issue every fourteen days, as the publication pattern states.
Your community's flagged items get worked as a queue against your written rules: each one read in full, decided on the rule it was flagged under, and logged with the action taken, with repeat patterns summed up at the end of a run so the rules can be tightened.
What changes
Honest fit
For you if: Your community's flagged items pile up and get settled by feel, so the same kind of item is decided two ways and nobody can show why.
Skip it if: You need the community rules written or actions carried out in the space on their own; the owner owns the rules, and the owner and their named moderators carry out the actions the log records.
Your first run produces: one real queue worked end to end against the written rules as they stand, in about forty minutes: the rules read in full with any gap noted, the flagged items counted before any are worked and the count written at the top of the log, the items worked oldest first with a category, a rule, a decision, and the action taken entered for each one, rule gaps and disputes set on the hold list for the owner, and a first short pattern note summing the run by category and by rule.
Preview, from the skill itself
“# Community Mod Queue<br># Community Mod Queue A community space collects flags the way a desk collects paper.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: 3837f5ed4f28… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 32.5 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/creator-community-evening-queue/WALKTHROUGH.md, examples/hobby-forum-monday-queue/WALKTHROUGH.md, reference/recording-each-decision.md, reference/spotting-repeat-patterns.md, reference/working-the-queue-in-order.md, reference/writing-the-rules-down.md, templates/pattern-report.template.md, templates/queue-log.template.md
Worked example, from the package
examples/creator-community-evening-queue/WALKTHROUGH.md
A community space attached to a small video channel, run by the channel owner with two moderators. The queue is worked each evening. This is the run on Thursday 4 June 2026, covering the items flagged that day.
Seasonal hiring stops running on hurry and memory: the role is written from the work the season actually holds, the screening questions and pass marks are fixed before the first application arrives, trial days are paid and recorded against the same checks for everyone, and every candidate, hired or not, gets an ending in writing.
What changes
Honest fit
For you if: You hire seasonal hands on hurry and memory, with a vague ad, different questions for different people, and silence for everyone not picked.
Skip it if: You need the role advertised, candidates contacted, or the hire decided; the operator places the ad, sends every note, and makes the call, and this pack keeps the record the decision stands on.
Your first run produces: the pack built for one role for the next season, end to end on paper before any ad goes out, in about an hour: last season's records listed week by week for the work the season actually held, the role written from that work with the must haves kept apart from what can be taught, the screening questions and pass marks fixed and dated, the trial day shape and its pay rate set, and the pack sheet laid out with a row per stage, ready to take its first application.
Preview, from the skill itself
“# Seasonal Crew Hiring Pack<br># Seasonal Crew Hiring Pack Seasonal hiring usually runs on memory and hurry.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: a02c1041495d… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 35.4 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/autumn-yard-crew-hiring-run/WALKTHROUGH.md, examples/spring-cleanup-hiring-run/WALKTHROUGH.md, reference/closing-out-offers-and-rejections.md, reference/screening-questions-fixed-in-advance.md, reference/trial-days-on-the-record.md, reference/writing-the-role-from-the-work.md, templates/hiring-pack-sheet.template.md, templates/trial-day-record.template.md
Worked example, from the package
examples/autumn-yard-crew-hiring-run/WALKTHROUGH.md
A garden and yard operator, Priya, works with one lead, Tomas, and hires one seasonal hand for the autumn season: about eight weeks of leaf clearing, hedge and garden tidies, and storm debris after bad weather. She needs the hand for weekdays and for Sunday cover one week in two, which is the requirement this run turned on.
Every cash payment is logged the moment it lands, with the job, the amount, and who took it, counted and deposited on a stated rhythm, and reconciled against the invoices it settles, so a shortage is named on the record the day it appears.
What changes
Honest fit
For you if: Your work takes cash on doorsteps and in driveways, and by the end of the week nobody can say which jobs it came from or whether all of it is still there.
Skip it if: You need banking done, accounts kept, or tax advice; the operator makes each deposit, and this log only records the counted figure and the deposit date the operator confirms.
Your first run produces: the log set up and the opening position entered from a real count, in about thirty minutes: the log opened from the current invoices and job records with a line ready for each payment, the count and deposit rhythm written at the top in plain words with the next count date named, any cash already on hand counted bill by bill and entered as the opening line dated today, and cash no record can tie to a job entered as its own line and named as unassigned.
Preview, from the skill itself
“# Cash Handling Log<br># Cash Handling Log Cash is the easiest money to lose track of.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: 08dba643dac0… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 37.6 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/clean-week-reconciled-in-full/WALKTHROUGH.md, examples/short-count-named-same-day/WALKTHROUGH.md, reference/counting-and-deposit-rhythm.md, reference/logging-cash-at-receipt.md, reference/reconciling-against-invoices.md, reference/shortages-and-discrepancies.md, templates/cash-log.template.md, templates/reconciliation-and-deposit.template.md
Worked example, from the package
examples/clean-week-reconciled-in-full/WALKTHROUGH.md
A fictional local service operator who cleans gutters, trims hedges, and repairs fences for householders. Two people take cash: Dan, the operator, and Luis, who works with him. Some customers pay by card or transfer. Some pay in cash on the day. The rhythm on the log is counted every Friday and deposited the same day. This is the week of Monday 6 April 2026 to Friday 10 April 2026.
Every guarantee you promise on a job sits in one ledger with what was promised, on which job, and until when, and every claim against a guarantee is tracked to its recorded resolution.
What changes
Honest fit
For you if: You give guarantees on jobs and months later nobody can say for sure what was promised, on which job, or until when.
Skip it if: You need guarantees given, terms set, coverage judged, or repair work done; those calls and that work stay with the operator, and this ledger keeps the record of promises and claims.
Your first run produces: the first ledger built from the job records, with any open claim logged in the same sitting, in about forty-five minutes: the recent job sheets and invoices gone through for each guarantee given in writing, one entry written per promise with the job, the wording used, what is not covered, and the end date worked from the start date and the term, each open claim logged against its entry with a next step named, and the next review date written at the top.
Preview, from the skill itself
“# Service Guarantee Tracking<br># Service Guarantee Tracking A guarantee is easy to give and hard to remember.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: 74306d1fd531… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 34.0 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/kitchen-and-bath-ledger-opening/WALKTHROUGH.md, examples/roofing-season-claims-review/WALKTHROUGH.md, reference/closing-a-claim-with-resolution.md, reference/recording-a-claim.md, reference/reviewing-the-ledger-and-expiry.md, reference/writing-a-guarantee-entry.md, templates/claim-record.template.md, templates/guarantee-ledger.template.md
Worked example, from the package
examples/kitchen-and-bath-ledger-opening/WALKTHROUGH.md
A fictional local firm, Millbrook Kitchen and Bath, fits kitchens and bathrooms in and around one town. Guarantees were given on paper with each job, but the papers sat in separate job files. On 2 June 2026 the operator opens one ledger and brings the recorded promises together. This walk shows the opening entries and the one claim open on the day.
Your job photos move from phone to archive on one fixed routine: each batch counted, copied, reconciled, and verified in the archive before the phone copy is cleared, so the only copy of a job's photos never lives on a phone.
What changes
Honest fit
For you if: Your job photos live on your phone, and the only copy of a finished job's photos is one lost or broken phone away from gone.
Skip it if: You need photos taken, filed against job records, or lost photos recovered; this routine starts after the photos exist and moves them to an archive you name, and it clears no phone copy on its own.
Your first run produces: the archive set up and one real backup run on the photos now on the phone, in about thirty minutes: the archive and its layout named with one folder per batch, the routine day written at the top of the log, the photos grouped into batches by job or named set, each batch counted on the phone, copied alone to its archive folder, counted again and reconciled, a sample of its files opened in the archive, and each batch marked passed or held, with no phone copy cleared yet.
Preview, from the skill itself
“# Photo Backup Routine<br># Photo Backup Routine A phone fills up with job photos, and a phone is a poor place for the only copy to live.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: 39cd9ff1ad59… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 34.8 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/mixed-jobs-discrepancy-held/WALKTHROUGH.md, examples/roofing-week-clean-backup/WALKTHROUGH.md, reference/choosing-the-routine-and-archive.md, reference/clearing-the-phone-safely.md, reference/counting-and-reconciling-batches.md, reference/verifying-the-archive-copy.md, templates/backup-log.template.md, templates/batch-verification-note.template.md
Worked example, from the package
examples/mixed-jobs-discrepancy-held/WALKTHROUGH.md
A local service operator who does fence work, deck cleaning, and tree work. Photos are backed up every Monday morning for the week just ended. The archive holds one folder per batch, named with the date and the job name. This is the run on Monday 11 May 2026, covering three jobs. All counts below are still photos. No batch in this run held videos.
A changed brief format or a repeating handoff fault stops being a guess: a fixed set of scripted cases runs the handoff cold, one plain pass rule decides it, and every failure is logged against the brief field that broke until the same case passes on retest.
What changes
Honest fit
For you if: Agents hand work to each other on a brief format, and the same kind of gap keeps surfacing after the handoff, in live work.
Skip it if: You need briefs written or real work handed over; this skill tests handoffs against the brief format and records where they fail, and it grades no agent and changes no format on its own.
Your first run produces: one real handoff tested as it stands, in about ninety minutes: the sending agent, the receiving agent, the work the handoff covers, and the brief format version named at the head of a run record, six cases built (a sound routine handoff, one with part done work in the open items, one with a decision the next agent must stand behind, one with a watch item that may land in the next run, one carrying a known gap written as an open flag, and one built from the last real handoff that went wrong), the expected result written on each case before any case is run, all six run cold in one sitting with the receiving side working from each brief alone, each case scored against the pass rule, and every failure written on the run record against its brief field, with the date of the next run set.
Preview, from the skill itself
“# Multi-Agent Handoff Tests<br># Multi-Agent Handoff Tests A handoff can read as complete and still leave the next agent short.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: 0bbe5992a436… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 53.9 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/eight-cases-for-a-draft-to-review-handoff/WALKTHROUGH.md, examples/six-cases-for-an-intake-to-research-handoff/WALKTHROUGH.md, reference/building-the-case-set.md, reference/handoff-test-terms.md, reference/recording-failures-against-the-brief.md, reference/the-pass-rule-and-scoring.md, templates/handoff-test-case.template.md, templates/handoff-test-run.template.md
Worked example, from the package
examples/eight-cases-for-a-draft-to-review-handoff/WALKTHROUGH.md
A two person support desk used two agents in sequence. A drafting agent wrote draft replies to customer notes during the day. A review agent read the drafts in the evening, checked each one against the records it rested on, and lined the sound ones up for the operator to send in the morning. The handoff between them was a brief in the desk's usual format. Drafts kept arriving back with the same two notes: a draft marked as resting on a price that had since changed, and a draft held for a customer whose open item named no holder, so neither side chased it.
You can say how many records are wrong and which rule fails most, before a customer or a payment run finds out: standing records are sampled on a cadence and checked against written quality rules, error rates are reported per rule, and each correction is routed to the record owner.
What changes
Honest fit
For you if: Your work rests on standing records, customer addresses, supplier files, or lists that drift as people move and details change.
Skip it if: You need every record read, records corrected for you, or the quality standards set; this check samples records that already exist and reports rates per rule, and the record owners make the corrections.
Your first run produces: the first spot check run on one standing file, in about an hour: the run record opened with the file named, the operator and the cadence written down, and the size of the full file recorded, a rule list of four to six rules written before any sample is drawn, each rule one a checker can settle from a named source, a first sample of 40 to 80 records drawn by the stated method and written down, each sampled record checked against its source, records that cannot be checked marked with their reason and never swapped, and the faults counted per rule with each correction routed to its record owner.
Preview, from the skill itself
“# Data Quality Spot Checks<br># Data Quality Spot Checks Every standing file drifts.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: 2e61f3c36c07… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 43.8 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/monthly-supplier-records-check/WALKTHROUGH.md, examples/quarterly-customer-address-check/WALKTHROUGH.md, reference/choosing-a-sample.md, reference/reading-error-rates.md, reference/routing-and-closing-corrections.md, reference/writing-quality-rules.md, templates/correction-note.template.md, templates/spot-check-run.template.md
Worked example, from the package
examples/monthly-supplier-records-check/WALKTHROUGH.md
A small trade supply firm kept one file of supplier records, used for ordering and for paying invoices. Faults came to light one at a time: a payment run held up by a record with no terms on it, a delivery sent to a site address the supplier had left a year before. The operator set a monthly spot check, on the first Monday of the month, and this is the March run, in figures. The firm and its figures are an illustration, built to show the method.
The rare share of a secret stops being a handover in a hurry: one named recipient and one stated purpose, single use or a stated expiry, the share logged by label only, and the reach revoked or the value rotated when the window closes, so no past share sits open unaccounted.
What changes
Honest fit
For you if: You sometimes have to give a helper or a tool a secret to get a job done, and afterwards nobody can say who still holds it or until when.
Skip it if: You need a value held, fetched, or handed over for you, or standing reach granted; this protocol governs the one off share itself, the operator shares the value in their own tools, and no value ever appears in its records.
Your first run produces: the shares already open accounted for and the share log opened, in about forty-five minutes: every share still believed open listed from records and from memory, each one marked shown, stated, or unknown with its reason, one log entry written for each open share by label only, a closing action and a closing date agreed for each entry on the spot, and the shares whose jobs are long finished closed during this first run, not carried forward as paperwork.
Preview, from the skill itself
“# Secret Sharing Protocol<br># Secret Sharing Protocol Most secrets are shared in a hurry.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: e7f0b19a5628… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 39.9 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/giving-a-tool-one-afternoon-of-reach/WALKTHROUGH.md, examples/one-off-share-to-a-covering-helper/WALKTHROUGH.md, reference/choosing-the-channel-and-the-window.md, reference/closing-the-share-revoke-or-rotate.md, reference/logging-a-share-by-label.md, reference/naming-the-recipient-and-the-use.md, templates/share-log-entry.template.md, templates/share-request-and-approval.template.md
Worked example, from the package
examples/giving-a-tool-one-afternoon-of-reach/WALKTHROUGH.md
A small support operation kept one inbox for client questions, watched by an agent that sorted each morning's mail into a digest for the operator, Leila. On Tuesday 14 April 2026 the sorting had been wrong for three mornings running, and the fault sat somewhere in how the agent read the inbox. A log review tool the operation was trying, named in the request as the LogLens review tool, could read the run logs and the inbox together and show where the reading went wrong, but only if it was given reach to the support inbox secret, live copy, for the run. This is that share, from request to closing.
A gap in uptime cover is found in the roster and not during an outage: every coverage window carries a named holder and a named backup, the handoff between holders and the contact order are written down, and the whole roster is reviewed on a cadence.
What changes
Honest fit
For you if: Your services need a person to answer when they go down, and cover at nights, weekends, or holidays rests on habit more than on a written roster.
Skip it if: You need services monitored, incidents graded, or outage messages written; this roster names who answers before any incident exists, and it runs no check, pages no one, and fixes no outage itself.
Your first run produces: the first roster built from the team and the services as they stand, in about an hour: the services that need a person to answer when they go down listed, the coverage windows fixed for the week with the time zone named once at the top, one entry per service with the holder and the backup for each window, the contact order written for each person, how each handoff is confirmed settled, and the full week checked for gaps and thin cover, so the roster sits held for the operator with its review date set.
Preview, from the skill itself
“# Uptime Incident Roster<br># Uptime Incident Roster A service rarely goes down at a tidy hour.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: ee6e67578d51… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 42.2 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/four-person-team-fixes-night-handoffs/WALKTHROUGH.md, examples/two-person-studio-finds-a-sunday-gap/WALKTHROUGH.md, reference/finding-gaps-and-thin-cover.md, reference/handoffs-and-contact-order.md, reference/review-cadence-and-triggers.md, reference/roster-terms-and-windows.md, templates/coverage-entry.template.md, templates/roster-record.template.md
Worked example, from the package
examples/four-person-team-fixes-night-handoffs/WALKTHROUGH.md
A fictional four person team ran a small learning platform for paying members: a member login and course pages, and a checkout where new members joined. The team was Ana Costa, who owned the service, Ben Okafor, Clara Lind, and Dev Patel. Daytime cover worked well. Night cover was where the roster, once written, showed its faults.
Readers stop landing in the wrong place: on a set cadence every link and cross-reference in the docs is followed to its target, dead or moved targets are listed per page, and each one is fixed in place or flagged with the replacement recorded beside it.
What changes
Honest fit
For you if: Your docs carry links and cross-references that readers follow, and pages move or get renamed as the docs grow.
Skip it if: You need the writing on a page judged, guides restructured, or anything published; this check tests where links land, not whether the landing page is good, and the operator makes every fix and publishing call.
Your first run produces: the first link check run on one guide set or one help section, in about an hour: the record opened with the docs named, the cadence set, and where the record lives written down, the page list written in the order a reader meets the pages, then page by page every link and cross-reference on the page written down in order, each one followed and marked as holding or listed as dead, moved, wrong target, or section missing, and each replacement target followed too before it is written down, so the fix rests on a landing that was seen.
Preview, from the skill itself
“# Docs Link Checking<br># Docs Link Checking A documentation page can read perfectly and still fail its reader in one click.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: ea3b0dcdfc94… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 38.8 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/quarterly-sweep-that-found-moved-guides/WALKTHROUGH.md, examples/release-week-check-after-a-docs-move/WALKTHROUGH.md, reference/cadence-and-coverage.md, reference/checking-a-page-step-by-step.md, reference/fix-or-flag-decisions.md, reference/link-kinds-and-targets.md, templates/link-check-record.template.md, templates/page-link-entry.template.md
Worked example, from the package
examples/quarterly-sweep-that-found-moved-guides/WALKTHROUGH.md
A small studio publishes the help docs for Slotwise, its fictional booking tool, as twelve pages that grew one at a time over two years. Nobody owned the links. Guides had been moved during two site tidies, and readers had started writing in to say a link in one page landed on a missing page or in the wrong guide. The operator set a quarterly cadence, built the first link record, and ran the first full sweep over one week. This is how that sweep ran and what it settled.
The same term stops being rendered two ways: one glossary record holds each term's approved rendering in each language and the context that decides between variants, and a new term enters the record only with the source it was admitted from.
What changes
Honest fit
For you if: Your material is translated into more than one language, and the same term comes back rendered in more than one way.
Skip it if: You need documents translated or a translation judged for how well it reads; this record keeps the approved renderings and the terms, and the people who translate do the translating.
Your first run produces: the first glossary record built from one named set of material, in about an hour: the record opened with the material it covers, the source language, each language covered, and the owner named, candidate terms gathered from that one set alone, repeats of the same source term merged into one entry each, any list that already exists brought in with where each of its terms came from noted where that can still be found, candidates with no source that can be pointed to held off the record as leads, and the approved rendering per language written on each entry.
Preview, from the skill itself
“# Translation Glossary Keeping<br># Translation Glossary Keeping Translated material drifts in a quiet way.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: ef5357eaaae0… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 55.6 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/cleaning-up-after-two-variants-spread/WALKTHROUGH.md, examples/first-glossary-for-a-plant-care-guide/WALKTHROUGH.md, reference/admitting-and-retiring-terms.md, reference/checking-a-batch-against-the-glossary.md, reference/choosing-between-variants.md, reference/glossary-terms-and-entry-parts.md, templates/glossary-entry.template.md, templates/glossary-record.template.md
Worked example, from the package
examples/cleaning-up-after-two-variants-spread/WALKTHROUGH.md
Maren Bay Tours is a fictional boat tour business with a small set of booking help pages, written in English and issued in German. A glossary record already existed, with 34 live terms. Over one season, pages were added by two translators working at different times, and three terms came back from a batch check with renderings that sat apart from the record. This is how the cleanup was worked, term by term, in counts.
Closed jobs stop teaching nothing: each one is compared estimate line against actual line on a set cadence, the lines that keep running over or under are named as patterns, and the price list or scoping note adjustments those patterns support are drafted for your approval.
What changes
Honest fit
For you if: You estimate jobs line by line, close them, and write the next estimate the same way, with nothing carried from the last jobs into the next price.
Skip it if: You need a profit figure per job, an estimate built for a new job, or a closed job repriced; this review stops at estimate against actual, and every adjustment faces forward, to the next estimate, and waits for your approval.
Your first run produces: the last four closed jobs reviewed line by line, in about forty minutes: one sheet opened per job with the estimate lines copied in exactly as given, the actual figure from each closed job record lined up beside its estimate line, each variance worked in the open and each job total checked against the sum of its line variances, the same line read across all four sheets, any line that moved the same way on at least three of the four jobs named as a pattern with its figures, and one adjustment drafted from the clearest pattern into the review log.
Preview, from the skill itself
“# Estimate vs Actual Review<br># Estimate vs Actual Review An estimate is a guess written before the work starts.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: 3105d7a14ad7… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 42.7 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/deck-repair-review-that-changed-the-scope/WALKTHROUGH.md, examples/fence-repairs-monthly-pattern-review/WALKTHROUGH.md, reference/lining-up-estimate-and-actual-lines.md, reference/naming-the-recurring-patterns.md, reference/running-the-review-cadence.md, reference/turning-the-review-into-adjustments.md, templates/estimate-vs-actual-sheet.template.md, templates/review-pattern-and-adjustment-log.template.md
Worked example, from the package
examples/deck-repair-review-that-changed-the-scope/WALKTHROUGH.md
A local service operator who does small outdoor repairs around houses. A customer, Dana Whitfield, had a price agreed to replace worn deck boards and rebuild the deck steps. The job closed, and at the weekly review the operator set the estimate against the actual, line by line. This example follows that one sheet, the lesson it held on its own, and the scoping change it drafted. The pattern across jobs is a later example's work; this one shows a single sheet done properly.
Every sponsorship sits in one commitment log, what you promised to give and what the event promised back, with dates and deliverables tracked on both sides, so the renewal is decided from what actually came back instead of from memory.
What changes
Honest fit
For you if: Your business sponsors local events, fairs, teams, or town days, and what was promised on each side lives in talks, messages, and memory.
Skip it if: You need a marketing plan, events chosen, or the money itself handled; this log records the commitment and what came back, it promises no return from a sponsorship, and the renewal call stays with you.
Your first run produces: the log opened with the sponsorships the business is carrying now, in about thirty minutes: each live sponsorship, and any that closed in the last season and may be asked to renew, entered once, the event and its dates written, what the business gives entered with each amount or valued item shown and the total worked in the open, what the event gives back entered item by item, the due dates on both sides written, anything resting on a talk entered as from a talk with the date of the talk, and each closed sponsorship marked for what actually arrived, in full, in part, or not arrived.
Preview, from the skill itself
“# Local Event Sponsorship Log<br># Local Event Sponsorship Log Category: Local Services / Marketing”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: e701834d4c0c… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 38.3 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/county-fair-banner-and-booth/WALKTHROUGH.md, examples/youth-baseball-season-not-renewed/WALKTHROUGH.md, reference/logging-what-actually-came-back.md, reference/making-the-renewal-decision-from-the-record.md, reference/recording-what-was-promised-each-way.md, reference/tracking-dates-and-deliverables.md, templates/renewal-review.template.md, templates/sponsorship-log.template.md
Worked example, from the package
examples/county-fair-banner-and-booth/WALKTHROUGH.md
A local services business in this example is Patel Flow Services, a fictional drain and plumbing service run by Samira Patel. The event is a fictional county fair, the Harvest County Fair, held 11 to 13 September 2026. This example follows one sponsorship from agreement to close, and a renewal that was agreed only after one part return was put in writing for the next year.
Rented equipment stops being double counted: every booking sits on one schedule with its window, pickup and return times, and the job it serves, and every change is checked against that schedule, so a clash is caught on the page before it is caught on the morning.
What changes
Honest fit
For you if: You rent equipment for jobs, and each booking lives in its own diary entry, message, or slip from the yard.
Skip it if: You need a rental booked, extended, or moved, a yard or a customer called, or a clash decided; those calls and that spending stay with you, and this schedule only puts the dates, the times, and the cost of each option in front of you.
Your first run produces: the next rental booking placed on one schedule and clash checked, in about thirty minutes: the booking read as the yard gave it, the item, the daily rate, and the window it was booked for, the job diary read for the job it serves and for any other job that wants the same item near those dates, the booking entered on the schedule with its pickup and return times, its cost worked as the daily rate times its rental days, and the clash check run against every other booking for that item, including any booking that only touches the first or the last day of the window.
Preview, from the skill itself
“# Equipment Rental Scheduling<br># Equipment Rental Scheduling A rented machine can only be in one place at a time.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: ae1591744418… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 36.3 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/chipper-extension-that-held/WALKTHROUGH.md, examples/mini-excavator-clash-when-a-job-moved/WALKTHROUGH.md, reference/booking-a-rental-window.md, reference/catching-and-working-a-clash.md, reference/extending-or-shortening-a-rental.md, reference/pickup-and-return-times.md, templates/booking-and-change-note.template.md, templates/rental-schedule.template.md
Worked example, from the package
examples/chipper-extension-that-held/WALKTHROUGH.md
A local operator who clears overgrown ground and cuts back trees and hedges. The operator rents a wood chipper for clearing jobs. This example follows one chipper booking in July, work that ran long by a day, and an extension that was checked against the next booking for the same chipper before it was agreed, and held because the gap after it still stood.
Each property gets the same winterization run every season, taps, pipes, heating, and gutters worked in order and marked as they stand, with findings and follow-ups written on the run record, so next season starts from this season's record instead of from memory.
What changes
Honest fit
For you if: You winterize the same properties each season, and what was found or left open last season lives in one person's memory.
Skip it if: You need repairs done, a promise that a property cannot freeze, or a full winter closing of an empty property; this run records the checks, and repairs are follow ups done under your own arrangements.
Your first run produces: one property given a full winterization run, in about forty-five minutes: the run record opened with the property block filled in from the operator record before the visit, any prior run record for the property read first and its open items taken along, the checklist worked from the top at the property, outdoor taps first, then exposed pipes, then the heating checks, then the gutters, each item marked as it stands at the place it was found, and a follow up list naming each open item with its timing, with no item claimed done that was not done.
Preview, from the skill itself
“# Winterization Checklist Runs<br># Winterization Checklist Runs Winter comes to the same properties every year, and the same places tend to need attention: the outdoor tap that was left with a hose on, the pipe run along an outside wall, the heating filter no one has looked at since spring, and the gutter that filled with leaves in October.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: 185c037eb5cd… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 47.0 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/single-family-house-first-run/WALKTHROUGH.md, examples/small-rental-duplex-with-open-items/WALKTHROUGH.md, reference/heating-and-gutter-checks.md, reference/recording-findings-and-follow-ups.md, reference/taps-pipes-and-freeze-risk-notes.md, reference/the-full-winterization-checklist.md, templates/follow-up-and-next-season-summary.template.md, templates/winterization-run-record.template.md
Worked example, from the package
examples/single-family-house-first-run/WALKTHROUGH.md
A local service operator who looks after houses for owners who live away part of the year. This example follows the first run at a three bedroom single family house, home of Dana Whitfield, who was at the property on the day. The house has a basement, a crawl space under the kitchen end, a gas furnace, and guttering on the front, back, and porch. This was the first run recorded under this skill, so there was no prior run record to read.
The photos that go out with a quote are chosen from the job's own record and labeled to the quote line each one supports, and anything the record does not back is held back with its reason logged, so the customer can match every photo to its line.
What changes
Honest fit
For you if: You send quotes that would be clearer with photos, and the photos are picked at sending time, from memory or from the nearest folder.
Skip it if: You need a quote built from photos, prices set, quote lines written, or photos taken or improved; this skill starts only once a quote exists, and attaching and sending stay with you.
Your first run produces: the next quote ready to send given a checked photo set, in about thirty minutes: the quote read as it stands, each line in order as worded, the job record for that job read and the photos it holds listed with their references, each candidate photo matched to the line it supports and labeled to that line, any photo the record does not back set aside with its reason, the attachment sheet drafted with a count that matches the photos named on it, and a selection log that accounts for every photo considered, the held back ones included.
Preview, from the skill itself
“# Quote Photo Attachments<br># Quote Photo Attachments A quote on its own asks the customer to picture the work.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: db5ceebd51af… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 34.0 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/deck-repair-quote-with-labeled-photos/WALKTHROUGH.md, examples/gutter-quote-with-a-photo-left-out/WALKTHROUGH.md, reference/checking-the-attachment-set-before-it-goes-out.md, reference/choosing-photos-from-the-job-record.md, reference/holding-back-photos-the-record-does-not-back.md, reference/labeling-each-photo-to-its-quote-line.md, templates/photo-selection-log.template.md, templates/quote-attachment-sheet.template.md
Worked example, from the package
examples/deck-repair-quote-with-labeled-photos/WALKTHROUGH.md
A local service operator who does small repairs around houses. A customer, Ellen Ward, asked for a price to repair the back deck at her house: boards to replace, one joist section to repair, and fixings to renew. The quote was written first, from a visit. This example follows what happened next: the photos for that quote were chosen from the job record, labeled to their lines, and checked, with one photo held back.
Captions stop quietly changing what a video says: every caption line is checked against the spoken words before posting, errors are listed per video and fixed in the caption file itself, and names and numbers get a second pass of their own.
What changes
Honest fit
For you if: You post short videos with captions, and the captions come from a tool or from a quick read before they go out.
Skip it if: You need captions made from scratch for a video that has none, captions translated, or the video edited or posted; this check works on an existing caption file, against the spoken words in the same video.
Your first run produces: the next short video for posting given a full caption check, in about twenty minutes: the video played through once with no caption in front of you, just to hear it whole, the check sheet then worked line by line with each caption line's time, what the caption says, and what was spoken noted, every difference listed on the error list, the second pass for names and numbers run even when the main check found none, each confirmed error fixed in the caption file itself, and the fixed lines played once more against the audio.
Preview, from the skill itself
“# Short Video Caption Checks<br># Short Video Caption Checks A caption that says the wrong name or the wrong number is not a small typo.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 9, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: bf78a117d253… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 9, 2026): no leaks found.
Bundle size: 34.1 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/kitchen-tip-name-and-time-fixed/WALKTHROUGH.md, examples/workshop-demo-price-and-place-fixed/WALKTHROUGH.md, reference/checking-captions-against-the-audio.md, reference/fixing-errors-in-the-caption-file.md, reference/listing-errors-per-video.md, reference/names-numbers-and-places-second-pass.md, templates/caption-check-sheet.template.md, templates/caption-error-list.template.md
Worked example, from the package
examples/kitchen-tip-name-and-time-fixed/WALKTHROUGH.md
A fictional creator, Priya Nair, who posts short kitchen tips filmed at her home counter. This example follows one short video, about beans for a weeknight soup, from its first caption check to the fixed file held for her to post. All names, places, and figures in this example are made up for it, and the figures are the figures spoken in the example itself.
Every agent in the fleet is judged on its own record: the cost for one period set beside the work that agent actually delivered, ending with one recorded call per agent, keep, narrow, or retire, with the reason in plain words.
What changes
Honest fit
For you if: You run a fleet of agents and cannot say, from a record, which ones earn their cost, or why an agent was kept, narrowed, or retired.
Skip it if: You need the ledger kept, an agent shut down, or a savings figure promised; this review reads the cost record and the work record and records the calls, and the operator makes each one.
Your first run produces: the most recent full period reviewed at the smallest honest size, in about an hour: the review record started from the fleet template, every agent that ran in the period listed from the fleet record and the cost record together, one entry per agent with its cost and its delivered work read from the records and any gap named, one draft call per agent with its reason, the operator confirming or changing each call, and the next review date set before you stop.
Preview, from the skill itself
“# Agent Fleet Cost Review<br># Agent Fleet Cost Review A fleet can look affordable in total and still carry agents that no longer earn their place.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 10, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: f7b1878ab97d… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 10, 2026): no leaks found.
Bundle size: 41.2 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/first-review-for-a-four-agent-fleet/WALKTHROUGH.md, examples/quarterly-review-that-narrowed-two-agents/WALKTHROUGH.md, reference/joining-costs-to-delivered-work.md, reference/review-evidence-and-gaps.md, reference/setting-the-review-cadence.md, reference/the-three-calls.md, templates/agent-cost-review-entry.template.md, templates/fleet-cost-review.template.md
Worked example, from the package
examples/first-review-for-a-four-agent-fleet/WALKTHROUGH.md
A small design studio had run four agents for five months and had never once set their costs beside their work. The owner could see the monthly total rising and could not say which agent was behind the rise, or whether the rise was buying anything. She ran the first review over one month, the most recent full month in her records, in one sitting. Every name and figure here is invented for the example.
The public launch day runs from one written runbook: the checks before the announcement closed on evidence, posts and pages landing in a set order, a named watcher on every surface, and rollback and correction lines approved before the first post goes out.
What changes
Honest fit
For you if: You have a product, an update, or a service going public on a set day, with posts and pages that have to land in a set order.
Skip it if: You need the product built, the release run, or the posts published for you; this runbook orders the public day and its watching, and people still put each item out through its normal route.
Your first run produces: the first runbook built before the launch day, in about ninety minutes: what is launching, the announcement time, and the day lead written at its head, the checks before the announcement listed, the order of posts and pages set one line per item, a named watcher for every surface, the rollback and correction lines drafted and approved by the operator, and the log opened before the day starts.
Preview, from the skill itself
“# Launch Day Runbook<br># Launch Day Runbook A launch day looks simple from the outside.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 10, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: 24d6ca1b1c0d… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 10, 2026): no leaks found.
Bundle size: 40.6 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/first-launch-for-a-small-tool/WALKTHROUGH.md, examples/launch-day-with-a-late-correction/WALKTHROUGH.md, reference/before-the-announcement-checks.md, reference/order-of-posts-and-pages.md, reference/rollback-and-correction-lines.md, reference/who-watches-what.md, templates/launch-day-log.template.md, templates/launch-day-runbook.template.md
Worked example, from the package
examples/first-launch-for-a-small-tool/WALKTHROUGH.md
A two person studio had built a small tool that sorts job photos into folders by date and site name. It was their first public launch. The product was ready a week ahead, the page was drafted, and the announcement was two posts and a listing. Without a plan, the day would have been run from a chat thread. This is how the first runbook was built, the day before the launch, and how the day ran from it.
Archived files are tested on a set cadence against the archive index: each one opened in full and its hash compared with its index entry, so every failure is listed per file and can be repaired from the surviving copy while a good copy still exists.
What changes
Honest fit
For you if: You hold an archive that is trusted without ever being tested, and you want faults found while a good copy still exists to repair from.
Skip it if: You need the archive index built, files repaired, or a backup held; this check tests files against the index and lists failures per file, and repairs are the operator's separate step.
Your first run produces: the first full check of a small archive run as it stands, in about an hour: the check log started with the archive name, the scope, and the cadence, files with no index entry or no hash listed at once as coverage gaps, each file in scope then opened in full and its fresh hash compared with its entry, every result written on the log as it goes, failures placed on the failure list with a next step each, and the next check date set before you stop.
Preview, from the skill itself
“# Archive Integrity Checks<br># Archive Integrity Checks An archive can hold a file for years and still lose it quietly.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 10, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: fd3d6456bdd3… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 10, 2026): no leaks found.
Bundle size: 41.5 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/first-check-for-a-small-office-archive/WALKTHROUGH.md, examples/quarterly-check-that-found-three-failures/WALKTHROUGH.md, reference/cadence-and-scope.md, reference/check-fields-and-statuses.md, reference/reading-the-hash-check.md, reference/repair-from-the-surviving-copy.md, templates/failure-list.template.md, templates/integrity-check-log.template.md
Worked example, from the package
examples/first-check-for-a-small-office-archive/WALKTHROUGH.md
A small office with four staff held an archive of scanned invoices and signed delivery notes, built up over three years and never tested as a whole. The files sat in one shared store, and the index, built when the archive was started, held an entry and a hash for most of them. The owner could open any single file on request, but nobody could say whether every copy still matched the copy that was filed. This is how the first check was run, in one afternoon, on a monthly cadence the office then kept.
Lapsed customers are reviewed one by one from their own job records, and a plain win-back note is drafted for each from what that record supports, held for your approval before anything is sent, with the answer recorded either way.
What changes
Honest fit
For you if: You run local service work and have past customers who went quiet after a job that went well, and you want to reach them from their records, not from memory.
Skip it if: You need notes sent for you, offers or discounts invented, or customers chased after a no; the operator approves and sends every note, and one note per customer is the whole of a round.
Your first run produces: the first round run in about thirty minutes: the lapsed line set and written at the top of the log, five past customers past that line taken from the records, each job record read in full, anyone with open work, an open complaint, or an open price question ruled out with the reason logged, one note drafted per customer who remains and all of them held, the operator approving, editing, or dropping each one, and a log entry opened for every customer considered.
Preview, from the skill itself
“# Customer Win-Back Notes<br># Customer Win-Back Notes Category: Local Services / Reputation”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 10, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: 71fbf95e3d28… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 10, 2026): no leaks found.
Bundle size: 30.9 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/hauling-customer-after-a-long-gap/WALKTHROUGH.md, examples/yard-cleanup-customer-who-went-quiet/WALKTHROUGH.md, reference/drafting-the-win-back-note.md, reference/finding-lapsed-customers.md, reference/holding-for-approval.md, reference/recording-the-answer.md, templates/win-back-log-entry.template.md, templates/win-back-note.template.md
Worked example, from the package
examples/hauling-customer-after-a-long-gap/WALKTHROUGH.md
Names and details in this example are fictional. Dana is a generic invented customer, and the job is an invented one, written to show the shape of the work.
The coming year runs against a written plan built from your own prior year records: a planned figure and a reason for every income and expense class, totals worked in the open, and set review dates for reading the plan against actuals while the year can still be steered.
What changes
Honest fit
For you if: You run local service work, you hold a prior year of income and expense records, and you want the coming year written down as figures with reasons before the year starts moving.
Skip it if: You hold no usable records for the prior year, or your money is held in another system you trust and can check; a plan built with no records behind it is guesswork with a tidy cover, and this skill will not build one.
Your first run produces: the first plan built from the prior year records in one sitting: the plan started with the business name, the plan year, and the year the figures rest on, the income and expense classes listed exactly as the records name them, one planned figure per class with its prior year actual and its reason, a flag on any class no record can support, the two totals and the planned difference worked in the open, and the review cadence set with its first date written.
Preview, from the skill itself
“# Year Plan Budget Setup<br># Year Plan Budget Setup A new year usually starts with the same money questions as the last one, answered from habit.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 10, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: 57d2b56fa39b… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 10, 2026): no leaks found.
Bundle size: 41.2 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/first-year-plan-for-a-small-crew/WALKTHROUGH.md, examples/second-year-plan-after-a-full-year-of-actuals/WALKTHROUGH.md, reference/carrying-classes-forward.md, reference/handling-classes-that-change.md, reference/reviewing-plan-against-actuals.md, reference/setting-planned-figures-with-reasons.md, templates/plan-review.template.md, templates/year-plan.template.md
Worked example, from the package
examples/first-year-plan-for-a-small-crew/WALKTHROUGH.md
All business names, classes, and figures in this example are fictional example numbers, written to show the shape of the work. They are not the figures of any real business.
A workflow is planned end to end on one page: each step matched to a skill the operator holds or to a named gap, and the gaps ordered by what blocks the most work, so the cover held and the gaps that matter first are both plain to see.
What changes
Honest fit
For you if: You plan or run a workflow and want to know, step by step, which steps a held skill covers and which steps have no cover.
Skip it if: You need a gap filled, a skill bought or installed, or the workflow run for you; this plan names and orders the gaps from your own records, and what happens about each one is the operator's call.
Your first run produces: one workflow planned in about forty five minutes: the plan started with the workflow name, its end point, and where your skill records live, the steps listed in order with any step hiding two kinds of work split, the held list built from your own records only, each step matched and marked covered, partly covered, or gap with its reason, and the gaps ordered by blocked work with the next check date set before you stop.
Preview, from the skill itself
“# Skill Stack Planner<br># Skill Stack Planner A workflow rarely fails all at once.”
Package makeup: 1 skill file · 2 worked examples · 4 references · 2 templates
Vetting record
Scanned Nov 10, 2026 with the studio Skill Vetting scanner (33 signal patterns across six areas: prompt injection, dangerous commands, data exfiltration paths, obfuscation, secrets handling, environment harvesting).
Bundle sha256: 586006291f01… Short form. The full hash lives in the registry and is verified on every gate run.
Result: 0 findings on 9 files.
Second scan (gitleaks 8.30.1, Nov 10, 2026): no leaks found.
Bundle size: 41.4 KB · Format: SKILL.md package
Catalog record: GhostBus Hub skills on GitHub First-party package by Ghost Developer Studio; the package itself is delivered with purchase.
Listing record: 8 of 8
What's inside
SKILL.md, examples/planning-a-monthly-report-workflow/WALKTHROUGH.md, examples/planning-coverage-for-a-client-onboarding-workflow/WALKTHROUGH.md, reference/keeping-the-plan-current.md, reference/matching-steps-to-skills.md, reference/ordering-gaps-by-blocked-work.md, reference/stack-plan-fields.md, templates/skill-stack-plan.template.md, templates/workflow-step-map.template.md
Worked example, from the package
examples/planning-a-monthly-report-workflow/WALKTHROUGH.md
This example is fictional. The workflow, the people, and the skill names are invented for this walkthrough and stand for generic kinds of work and generic kinds of skill. Nothing here describes a real business or a real held set.
Every listing is scored against the same eight elements. Each one can be checked on the listing itself.
Every listing shows its own number. A missing element lowers the number and is named on the listing. The record is part of the listing, not a badge bought or granted.
Skills carry badges of their own, separate from member badges. A skill badge is computed from the skill's own listing record. It is not awarded and not kept: if a listing loses the thing a badge stands for, the badge goes away. A seller cannot buy a skill badge or give one to their own skill. Member badges are a different system: those are earned by people and kept.
The gate, named. These are the fourteen suites whose 135 checks run on every gate pass:
A second scan, from a tool we do not make. Every package is also run through gitleaks, an open-source secrets scanner, as a cross-check on the in-house scan. The most recent run, on Nov 4, 2026, found no secrets in any package. Two tools agreeing is stronger than one, and where they ever disagree, the listing will say so.
The honest limit: a scan covers known patterns. It is evidence, not a guarantee. Each listing names its files, so buyers can read exactly what they are getting.
The same procedure and the file lists are also published in that repo's README and catalog, so the check needs nothing from the Hub.
The Hub's working app already runs this check in one step, and it opens to buyers at launch. Until then, the procedure above needs nothing but the public catalog.
Questions about a skill, or one you want built? Email [email protected].
Phase 2 opens the Hub to outside sellers at launch, and the system behind it is already built and proven locally. It stays manual and curated at first, so quality stays high with no payout automation to get wrong on day one. Here is how selling works, step by step.
Sellers register and accept the seller terms before submitting anything, with a payout method recorded. Curation is manual: every seller is approved by the operator. The seller terms are published as version v1 on the Terms page.
A skill id belongs to the seller who listed it first. Nobody else can submit or update that id, and updates need a strictly higher version number, enforced at approval.
Every submission runs through the Skill Vetting scanner before a human reviews it. Any high severity signal holds the package, and a held submission can never be listed.
Approval by a human is the only way a submission becomes a listing. Listings carry your name, version, vetted badge, and bundle hash. Paid bundles live only in the private vault, exactly like the studio's own paid skills.
Every sale splits through a payout ledger in exact cents. The commission was decided on Nov 4, 2026: the Hub keeps 20% of the sale price and the seller receives 80%. Payouts are manual in this phase; Stripe Connect is a later step. The full seller terms are on the Terms page.
If a buyer is refunded, the seller's still owed payout reverses in the ledger with the reason recorded and history preserved. A payout that was already paid cannot be reversed in the ledger; that stays a manual step.
The split at the prices in the catalog today:
| Sale price | You keep 80% | The Hub keeps 20% |
|---|---|---|
| $12 | $9.60 | $2.40 |
| $15 | $12.00 | $3.00 |
| $19 | $15.20 | $3.80 |
| $20 | $16.00 | $4.00 |
Stripe fees come out of the Hub share, not yours (see Terms).
Every Hub listing carries the same public elements, whether the skill is the studio's own or a seller's. If your skill lists, its listing will show:
Every submission goes through the same five steps published in How vetting works on the What vetted means page: a dual scan of the package, a human review of the findings, a dated record published on the listing, the full file list published with the listing, and a bundle hash a buyer can recompute from the delivered files. There is no separate or lighter track for outside sellers, and the studio's own packages pass the same gate. A skill that fails vetting is not listed.
Questions about selling on the Hub? Email [email protected].
Humans browse here. Agents work through the registry underneath, using the bus tools they already have. Search, approval, install, and the audit trail can all happen inside GhostBus.
The agent searches the Hub registry workspace on GhostBus and finds the skill, its version, price, and hash. No human needs to go hunting for a download link. Free skills pull straight down.
The pull is written to the GhostBus provenance log, and file history on the bus tracks new versions. You can see what was pulled, when, and which hash was verified.
Each listing carries a bundle hash in the registry. On pull, the bundle's sha256 is checked against that hash. A bundle that has been changed does not match, and the pull does not pass as the listed version. License keys for paid skills are checked the same careful way: the Hub stores the hash only, never the key itself.
GhostMail contact cards live on the mail bus registry, where agents discover each other's mail addresses, logins, and ports. Phone extensions join the same contact cards as they arrive, so one registry answers both how to mail an agent and how to call one.
A paid pull starts as a purchase task with needsApproval: true. A claim before approval is refused. After human approval, the agent claims the task, pulls the bundle from the vault, and the hash is checked again.
Sending mail through GhostMail is approval gated, and there is no ungated send path. The same rule is being built for outbound calls on GhostPhone: drafted first, approved, then sent or dialed.
The full registry with complete hashes is verified on every run of the Hub's 135-check gate, and checkout opens at launch, with purchases agent-approved like everything else on the Hub.
Catalog: 200 skills, Nov 10, 2026.
{
"catalog": [
{
"id": "agent-seatbelt-guardrails",
"version": "0.3.2",
"price_usd": 0,
"category": "Security / Guardrails",
"bundle_hash_short": "a469a2603bce",
"bundle_kb": 10.0,
"composition": {"root": 1, "examples": 1},
"example_file": "examples/first-run/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": true,
"project_repository": "https://github.com/littlestjames82-sys/agent-seatbelt"
}
},
{
"id": "seatbelt-policy-authoring",
"version": "1.3.0",
"price_usd": 19,
"category": "Security / DevOps",
"bundle_hash_short": "da74e0f70ddf",
"bundle_kb": 107.3,
"composition": {"root": 1, "examples": 20, "reference": 9, "templates": 5},
"example_file": "examples/policy-audit/AUDIT.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false,
"project_repository": "https://github.com/littlestjames82-sys/agent-seatbelt"
}
},
{
"id": "skill-vetting",
"version": "1.0.0",
"price_usd": 15,
"category": "Agent Skills / Vetting",
"bundle_hash_short": "8ca6bc269c6d",
"bundle_kb": 50.5,
"composition": {"root": 1, "examples": 7, "reference": 2, "scripts": 1},
"example_file": "examples/clean-skill/AUDIT.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "skill-authoring",
"version": "1.0.0",
"price_usd": 19,
"category": "Agent Skills / Authoring",
"bundle_hash_short": "00fddc137435",
"bundle_kb": 30.9,
"composition": {"root": 1, "examples": 3, "reference": 6, "templates": 3},
"example_file": "examples/from-idea-to-listing/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "agent-mail",
"version": "1.0.0",
"price_usd": 15,
"category": "Agent Comms / Email",
"bundle_hash_short": "81c3849041cf",
"bundle_kb": 19.8,
"composition": {"root": 1, "examples": 2, "reference": 5, "templates": 2},
"example_file": "examples/license-delivery/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "agent-phone",
"version": "1.0.0",
"price_usd": 15,
"category": "Agent Comms / Phone",
"bundle_hash_short": "c7a1c0948144",
"bundle_kb": 19.3,
"composition": {"root": 1, "examples": 2, "reference": 5, "templates": 2},
"example_file": "examples/outbound-call/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "agent-incident-review",
"version": "1.0.0",
"price_usd": 12,
"category": "Agent Ops / Incident Review",
"bundle_hash_short": "d6cb3fd3b032",
"bundle_kb": 19.3,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/injection-attempt-review/REVIEW.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "agent-task-handoff",
"version": "1.0.0",
"price_usd": 15,
"category": "Agent Ops / Handoff",
"bundle_hash_short": "e616a99f0187",
"bundle_kb": 18.9,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/bus-handoff/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "release-gatekeeper",
"version": "1.0.0",
"price_usd": 15,
"category": "Agent Ops / Release",
"bundle_hash_short": "a45760b02556",
"bundle_kb": 17.9,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/version-truth/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "agent-memory-keeping",
"version": "1.0.0",
"price_usd": 15,
"category": "Agent Ops / Memory",
"bundle_hash_short": "b92354a292d5",
"bundle_kb": 19.2,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/correction-propagation/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "backups-restore-drills",
"version": "1.0.0",
"price_usd": 12,
"category": "Agent Ops / Backups",
"bundle_hash_short": "2e38fe6cfce1",
"bundle_kb": 18.4,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/hub-restore-drill/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "marketplace-replies",
"version": "1.0.0",
"price_usd": 12,
"category": "Agent Comms / Marketplace",
"bundle_hash_short": "7fced4111374",
"bundle_kb": 18.4,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/firewood-half-load/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "photo-job-quoting",
"version": "1.0.0",
"price_usd": 19,
"category": "Local Services / Quoting",
"bundle_hash_short": "e5c8a7af12e0",
"bundle_kb": 18.2,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/cleanup-lot-from-photos/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "load-safety-photo-check",
"version": "1.0.0",
"price_usd": 12,
"category": "Transport / Load Safety",
"bundle_hash_short": "ee561ea0011d",
"bundle_kb": 19.2,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/rear-overhang/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "agent-daily-briefing",
"version": "1.0.0",
"price_usd": 12,
"category": "Agent Ops / Briefing",
"bundle_hash_short": "01a286c5a37b",
"bundle_kb": 25.7,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/a-morning-with-signal/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "verified-research",
"version": "1.0.0",
"price_usd": 15,
"category": "Agent Ops / Research",
"bundle_hash_short": "eae2f35be458",
"bundle_kb": 26.4,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/the-land-record/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "job-site-scoping",
"version": "1.0.0",
"price_usd": 15,
"category": "Local Services / Scoping",
"bundle_hash_short": "1f7d43bd52b2",
"bundle_kb": 28.2,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/shed-concrete-prep/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "agent-watchkeeping",
"version": "1.0.0",
"price_usd": 12,
"category": "Agent Ops / Watches",
"bundle_hash_short": "3dec99b45c4d",
"bundle_kb": 30.7,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/the-reply-watch/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "context-cost-control",
"version": "1.0.0",
"price_usd": 15,
"category": "Agent Ops / Context",
"bundle_hash_short": "01c7673f0e43",
"bundle_kb": 19.9,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/the-oversized-brief/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "delegated-work-control",
"version": "1.0.0",
"price_usd": 19,
"category": "Agent Ops / Delegation",
"bundle_hash_short": "a334f734ac13",
"bundle_kb": 20.2,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/the-clean-handoff/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "invoice-and-follow-up",
"version": "1.0.0",
"price_usd": 15,
"category": "Local Services / Money",
"bundle_hash_short": "4ca84478d9e5",
"bundle_kb": 21.6,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/balance-after-deposit/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "review-requests",
"version": "1.0.0",
"price_usd": 12,
"category": "Local Services / Reputation",
"bundle_hash_short": "377e4a9ce1ee",
"bundle_kb": 21.4,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/the-fix-first-job/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "social-posting-gate",
"version": "1.0.0",
"price_usd": 15,
"category": "Studio / Publishing",
"bundle_hash_short": "e183ec88115b",
"bundle_kb": 22.0,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/the-almost-claim/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "lead-intake",
"version": "1.0.0",
"price_usd": 15,
"category": "Local Services / Sales",
"bundle_hash_short": "1cff687cf5fe",
"bundle_kb": 28.9,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/the-complete-referral/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "job-estimate-builder",
"version": "1.0.0",
"price_usd": 19,
"category": "Local Services / Money",
"bundle_hash_short": "9084189437d7",
"bundle_kb": 29.5,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/the-price-that-could-not-be-sourced/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "changelog-keeper",
"version": "1.0.0",
"price_usd": 12,
"category": "Studio / Shipping",
"bundle_hash_short": "6e87d4d56052",
"bundle_kb": 24.0,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/the-messy-range/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "agent-onboarding",
"version": "1.0.0",
"price_usd": 15,
"category": "Agent Operations",
"bundle_hash_short": "621872e31358",
"bundle_kb": 39.6,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/research-agent-for-a-small-studio/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "task-queue-triage",
"version": "1.0.0",
"price_usd": 12,
"category": "Agent Operations",
"bundle_hash_short": "d09f323584a2",
"bundle_kb": 36.8,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/agent-backlog-busy-week/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "agent-retirement",
"version": "1.0.0",
"price_usd": 12,
"category": "Agent Operations",
"bundle_hash_short": "51912acfa03a",
"bundle_kb": 34.4,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/emergency-misbehaving-agent/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "error-budget-watch",
"version": "1.0.0",
"price_usd": 12,
"category": "Agent Operations",
"bundle_hash_short": "ebefeb2d0a7a",
"bundle_kb": 33.8,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/budget-breach-and-narrowing/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "permission-audit",
"version": "1.0.0",
"price_usd": 15,
"category": "Agent Safety and Governance",
"bundle_hash_short": "6a3baf6a68de",
"bundle_kb": 27.8,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/before-a-bigger-role/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "secrets-hygiene",
"version": "1.0.0",
"price_usd": 19,
"category": "Agent Safety and Governance",
"bundle_hash_short": "dc6c1ca609bb",
"bundle_kb": 34.5,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/new-project-day-one/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "quote-follow-up",
"version": "1.0.0",
"price_usd": 12,
"category": "Local Services / Sales",
"bundle_hash_short": "2e6bd16055ce",
"bundle_kb": 25.8,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/clean-no-closeout/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "job-photo-documentation",
"version": "1.0.0",
"price_usd": 12,
"category": "Local Services / Field Operations",
"bundle_hash_short": "237d50513845",
"bundle_kb": 31.6,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/cleanup-before-set-settles-it/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "materials-run-planner",
"version": "1.0.0",
"price_usd": 15,
"category": "Local Services / Field Operations",
"bundle_hash_short": "7cc0144c47c8",
"bundle_kb": 33.5,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/repair-job-single-morning-run/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "customer-update-messages",
"version": "1.0.0",
"price_usd": 12,
"category": "Local Services / Reputation",
"bundle_hash_short": "5286ab1885a4",
"bundle_kb": 29.2,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/full-day-two-jobs/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "warranty-and-callbacks",
"version": "1.0.0",
"price_usd": 15,
"category": "Local Services / Field Operations",
"bundle_hash_short": "714a15dada8a",
"bundle_kb": 29.5,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/callback-outside-promise-priced-fix/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "seasonal-service-offers",
"version": "1.0.0",
"price_usd": 12,
"category": "Local Services / Marketing",
"bundle_hash_short": "e21925b126f4",
"bundle_kb": 27.7,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/firewood-push-first-cold-snap/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "agent-health-check",
"version": "1.0.0",
"price_usd": 15,
"category": "Agent Operations",
"bundle_hash_short": "524fb2e73772",
"bundle_kb": 40.0,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/the-drifting-inbox-agent/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "agent-cost-ledger",
"version": "1.0.0",
"price_usd": 15,
"category": "Agent Operations",
"bundle_hash_short": "17a856dfe5a9",
"bundle_kb": 31.5,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/the-month-end-close/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "agent-red-team-drills",
"version": "1.0.0",
"price_usd": 19,
"category": "Agent Safety and Governance",
"bundle_hash_short": "683b8ef47589",
"bundle_kb": 46.9,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/the-planted-note-drill.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "spreadsheet-cleanup",
"version": "1.0.0",
"price_usd": 12,
"category": "Data, Files, and Research",
"bundle_hash_short": "8aa330446e7d",
"bundle_kb": 34.0,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/event-signup-sheet-cleanup/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "document-intake-filing",
"version": "1.0.0",
"price_usd": 15,
"category": "Data, Files, and Research",
"bundle_hash_short": "f0022e1b827d",
"bundle_kb": 33.8,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/statements-and-a-form.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "file-filing-and-renaming",
"version": "1.0.0",
"price_usd": 12,
"category": "Data, Files, and Research",
"bundle_hash_short": "749ffd6958a6",
"bundle_kb": 38.6,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/the-client-files-sort/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "video-shorts-pipeline",
"version": "1.0.0",
"price_usd": 15,
"category": "Creator and Audience",
"bundle_hash_short": "f16a1f15a209",
"bundle_kb": 33.5,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/interview-to-shorts/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "audience-comment-triage",
"version": "1.0.0",
"price_usd": 12,
"category": "Creator and Audience",
"bundle_hash_short": "ea0c70ab1a3c",
"bundle_kb": 30.6,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/dm-backlog-after-a-live-session/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "referral-requests",
"version": "1.0.0",
"price_usd": 12,
"category": "Local Services / Reputation",
"bundle_hash_short": "ec1581e94d56",
"bundle_kb": 33.5,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/firewood-regular-names-a-friend/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "crew-day-planner",
"version": "1.0.0",
"price_usd": 15,
"category": "Local Services / Field Operations",
"bundle_hash_short": "ebaaef864ec7",
"bundle_kb": 38.6,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/four-jobs-two-crews-split-day/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "expense-and-receipt-log",
"version": "1.0.0",
"price_usd": 12,
"category": "Local Services / Money",
"bundle_hash_short": "dea382ffc913",
"bundle_kb": 30.1,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/valley-roof-repair/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "newsletter-issue-builder",
"version": "1.0.0",
"price_usd": 15,
"category": "Studio / Publishing",
"bundle_hash_short": "3e5990956456",
"bundle_kb": 33.8,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/quiet-week-client-work-roundup/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "meeting-notes-to-actions",
"version": "1.0.0",
"price_usd": 12,
"category": "Agent Operations",
"bundle_hash_short": "b5a27c4ddfdd",
"bundle_kb": 42.8,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/vendor-kickoff-walkthrough.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "release-announcements",
"version": "1.0.0",
"price_usd": 12,
"category": "Studio / Publishing",
"bundle_hash_short": "095d2e7fce0c",
"bundle_kb": 32.8,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/the-held-announcement-set.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "dependency-update-watch",
"version": "1.0.0",
"price_usd": 12,
"category": "Studio / Shipping",
"bundle_hash_short": "c2ebf204a30d",
"bundle_kb": 34.6,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/the-advisory-led-pass.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "data-handling-rules",
"version": "1.0.0",
"price_usd": 15,
"category": "Agent Safety and Governance",
"bundle_hash_short": "83460cf11374",
"bundle_kb": 36.0,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/the-client-file-handover/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "tool-output-verification",
"version": "1.0.0",
"price_usd": 15,
"category": "Agent Safety and Governance",
"bundle_hash_short": "c33191fe57db",
"bundle_kb": 37.8,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/the-doubled-price/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "source-archiving",
"version": "1.0.0",
"price_usd": 12,
"category": "Data, Files, and Research",
"bundle_hash_short": "27fe8a9eea88",
"bundle_kb": 32.1,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/the-price-page-behind-an-estimate.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "service-menu-builder",
"version": "1.0.0",
"price_usd": 15,
"category": "Local Services / Marketing",
"bundle_hash_short": "5c26c21e1800",
"bundle_kb": 40.4,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/hauling-and-cleanup-menu/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "job-completion-handover",
"version": "1.0.0",
"price_usd": 12,
"category": "Local Services / Field Operations",
"bundle_hash_short": "6aabbe5c724b",
"bundle_kb": 35.3,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/bathroom-refit-handover/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "weather-reschedule-planner",
"version": "1.0.0",
"price_usd": 12,
"category": "Local Services / Field Operations",
"bundle_hash_short": "0b2a67af147a",
"bundle_kb": 42.0,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/cold-snap-paint-week/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "customer-check-in-messages",
"version": "1.0.0",
"price_usd": 12,
"category": "Local Services / Reputation",
"bundle_hash_short": "c94c20b62eb4",
"bundle_kb": 36.1,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/furnace-filter-winter-check-in/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "podcast-episode-package",
"version": "1.0.0",
"price_usd": 15,
"category": "Creator and Audience",
"bundle_hash_short": "43adc7c7a3de",
"bundle_kb": 30.3,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/the-bakery-interview/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "deposit-terms-setup",
"version": "1.0.0",
"price_usd": 12,
"category": "Local Services / Money",
"bundle_hash_short": "2c677f6ade81",
"bundle_kb": 32.8,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/the-fence-job-materials-upfront/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "agent-escalation-handling",
"version": "1.0.0",
"price_usd": 15,
"category": "Agent Operations",
"bundle_hash_short": "73cff75ad35a",
"bundle_kb": 38.5,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/the-invoice-over-the-line/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "multi-agent-job-split",
"version": "1.0.0",
"price_usd": 19,
"category": "Agent Operations",
"bundle_hash_short": "871a9051b451",
"bundle_kb": 36.3,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/the-help-center-rebuild/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "privacy-consent-records",
"version": "1.0.0",
"price_usd": 15,
"category": "Agent Safety and Governance",
"bundle_hash_short": "a8e2d6442119",
"bundle_kb": 34.1,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/the-new-client-intake.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "uptime-and-status-checks",
"version": "1.0.0",
"price_usd": 12,
"category": "Studio / Shipping",
"bundle_hash_short": "fda0bb2592a5",
"bundle_kb": 34.8,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/the-morning-after-a-deploy/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "user-guide-writing",
"version": "1.0.0",
"price_usd": 15,
"category": "Studio / Publishing",
"bundle_hash_short": "47c53c4cafeb",
"bundle_kb": 33.6,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/repairdesk-first-guide.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "research-comparison-table",
"version": "1.0.0",
"price_usd": 12,
"category": "Data, Files, and Research",
"bundle_hash_short": "bb0bf11b20d8",
"bundle_kb": 40.6,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/backup-tools-for-a-small-studio.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "job-profit-review",
"version": "1.0.0",
"price_usd": 15,
"category": "Local Services / Money",
"bundle_hash_short": "e7b3cbdf0135",
"bundle_kb": 34.3,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/gravel-driveway-regrade/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "subcontractor-coordination",
"version": "1.0.0",
"price_usd": 15,
"category": "Local Services / Field Operations",
"bundle_hash_short": "78c1542d0747",
"bundle_kb": 35.0,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/bathroom-tile-plumber-sub/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "equipment-maintenance-log",
"version": "1.0.0",
"price_usd": 12,
"category": "Local Services / Field Operations",
"bundle_hash_short": "371af5b8bacc",
"bundle_kb": 32.1,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/log-splitter-first-season-on-record.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "google-profile-updates",
"version": "1.0.0",
"price_usd": 12,
"category": "Local Services / Marketing",
"bundle_hash_short": "d6a05c16da20",
"bundle_kb": 29.1,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/holiday-hours-roofing-update/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "short-form-topic-bank",
"version": "1.0.0",
"price_usd": 12,
"category": "Creator and Audience",
"bundle_hash_short": "22262b950e15",
"bundle_kb": 30.2,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/repair-crew-question-bank.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "new-customer-welcome-pack",
"version": "1.0.0",
"price_usd": 12,
"category": "Local Services / Sales",
"bundle_hash_short": "c9c356cc33c4",
"bundle_kb": 34.3,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/roof-valley-repair-first-job/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "agent-config-versioning",
"version": "1.0.0",
"price_usd": 15,
"category": "Agent Operations",
"bundle_hash_short": "26ca48fb602c",
"bundle_kb": 41.4,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/the-broken-briefing/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "scheduled-report-builder",
"version": "1.0.0",
"price_usd": 12,
"category": "Agent Operations",
"bundle_hash_short": "1e585d8790ad",
"bundle_kb": 36.4,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/the-monthly-cost-roundup/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "secrets-in-transit-check",
"version": "1.0.0",
"price_usd": 15,
"category": "Agent Safety and Governance",
"bundle_hash_short": "97076d088b60",
"bundle_kb": 38.8,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/the-debug-handoff/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "incident-comms-drafts",
"version": "1.0.0",
"price_usd": 12,
"category": "Studio / Publishing",
"bundle_hash_short": "ea5430dec6b2",
"bundle_kb": 38.3,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/checkout-failure-live-updates.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "beta-feedback-loop",
"version": "1.0.0",
"price_usd": 15,
"category": "Studio / Shipping",
"bundle_hash_short": "784a948cb872",
"bundle_kb": 39.8,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/the-duplicate-wave/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "csv-to-database-import",
"version": "1.0.0",
"price_usd": 15,
"category": "Data, Files, and Research",
"bundle_hash_short": "50e1891d876d",
"bundle_kb": 33.5,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/member-list-import/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "pre-job-site-checklist",
"version": "1.0.0",
"price_usd": 12,
"category": "Local Services / Field Operations",
"bundle_hash_short": "c745e67aacfe",
"bundle_kb": 34.1,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/fence-and-gate-day-before/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "service-area-pricing-zones",
"version": "1.0.0",
"price_usd": 15,
"category": "Local Services / Money",
"bundle_hash_short": "fb1d3cc60a2c",
"bundle_kb": 33.9,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/cleaning-operator-named-zones/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "complaint-handling-ladder",
"version": "1.0.0",
"price_usd": 15,
"category": "Local Services / Reputation",
"bundle_hash_short": "85594ab9676f",
"bundle_kb": 33.6,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/free-redo-asked-beyond-warranty/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "before-after-gallery-keeping",
"version": "1.0.0",
"price_usd": 12,
"category": "Local Services / Marketing",
"bundle_hash_short": "58a0f7968179",
"bundle_kb": 27.5,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/deck-cleaning-pair-cleared-for-use/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "series-and-sequels-planner",
"version": "1.0.0",
"price_usd": 12,
"category": "Creator and Audience",
"bundle_hash_short": "fc399db0d725",
"bundle_kb": 30.7,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/raised-bed-garden-series/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "sponsorship-outreach-drafts",
"version": "1.0.0",
"price_usd": 15,
"category": "Creator and Audience",
"bundle_hash_short": "134671316d45",
"bundle_kb": 28.9,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/cooking-channel-return-round/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "agent-dry-run-rehearsal",
"version": "1.0.0",
"price_usd": 15,
"category": "Agent Operations",
"bundle_hash_short": "dc7f40f1c6c7",
"bundle_kb": 36.5,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/the-invoice-chase-rehearsal/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "task-time-estimates",
"version": "1.0.0",
"price_usd": 12,
"category": "Agent Operations",
"bundle_hash_short": "2744db7b1013",
"bundle_kb": 34.7,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/the-inbox-that-always-ran-long/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "output-format-contracts",
"version": "1.0.0",
"price_usd": 15,
"category": "Agent Safety and Governance",
"bundle_hash_short": "e5b345ec340e",
"bundle_kb": 33.5,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/the-invoice-handoff/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "vendor-and-tool-renewals",
"version": "1.0.0",
"price_usd": 12,
"category": "Studio / Shipping",
"bundle_hash_short": "de5189866fd4",
"bundle_kb": 40.9,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/the-look-ahead-review.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "doc-drift-checks",
"version": "1.0.0",
"price_usd": 12,
"category": "Studio / Publishing",
"bundle_hash_short": "a35b95fb113f",
"bundle_kb": 44.3,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/guide-drift-after-release/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "pdf-form-filling",
"version": "1.0.0",
"price_usd": 15,
"category": "Data, Files, and Research",
"bundle_hash_short": "3be693493d71",
"bundle_kb": 40.2,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/event-permit-application.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "delivery-and-dropoff-runs",
"version": "1.0.0",
"price_usd": 12,
"category": "Local Services / Field Operations",
"bundle_hash_short": "b88031f6da78",
"bundle_kb": 29.5,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/five-stop-round-missed-delivery/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "rental-and-borrowed-gear-log",
"version": "1.0.0",
"price_usd": 12,
"category": "Local Services / Field Operations",
"bundle_hash_short": "6ae5651d82ca",
"bundle_kb": 36.4,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/borrowed-scaffold-tower-house-job.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "receipts-reconciliation",
"version": "1.0.0",
"price_usd": 15,
"category": "Local Services / Money",
"bundle_hash_short": "90dd40283505",
"bundle_kb": 33.7,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/five-jobs-one-short-payment.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "referral-partner-list",
"version": "1.0.0",
"price_usd": 12,
"category": "Local Services / Marketing",
"bundle_hash_short": "e8150d6938e8",
"bundle_kb": 38.8,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/cleaner-agent-circle-and-a-quiet-partner/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "live-session-prep",
"version": "1.0.0",
"price_usd": 15,
"category": "Creator and Audience",
"bundle_hash_short": "e1ff5d1a1a51",
"bundle_kb": 33.1,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/sketching-tutorial-live/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "comment-to-content-ideas",
"version": "1.0.0",
"price_usd": 12,
"category": "Creator and Audience",
"bundle_hash_short": "ffbf204e0cc0",
"bundle_kb": 33.4,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/beginner-baking-follow-ups/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "model-and-tool-swap-checks",
"version": "1.0.0",
"price_usd": 15,
"category": "Agent Operations",
"bundle_hash_short": "68814a48c0cc",
"bundle_kb": 41.6,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/the-model-upgrade/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "standing-instructions-review",
"version": "1.0.0",
"price_usd": 12,
"category": "Agent Operations",
"bundle_hash_short": "caa3e32f92e9",
"bundle_kb": 44.5,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/the-conflicting-reply-rules/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "data-retention-schedule",
"version": "1.0.0",
"price_usd": 15,
"category": "Agent Safety and Governance",
"bundle_hash_short": "1a335c91f371",
"bundle_kb": 33.4,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/first-schedule-cleanup/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "public-corrections-log",
"version": "1.0.0",
"price_usd": 12,
"category": "Studio / Publishing",
"bundle_hash_short": "435d9ec37ec1",
"bundle_kb": 37.1,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/the-wrong-date/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "archive-and-sunset-products",
"version": "1.0.0",
"price_usd": 15,
"category": "Studio / Shipping",
"bundle_hash_short": "0306a635698e",
"bundle_kb": 42.5,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/the-clean-retirement/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "spreadsheet-to-report",
"version": "1.0.0",
"price_usd": 12,
"category": "Data, Files, and Research",
"bundle_hash_short": "f82a3a3e5263",
"bundle_kb": 33.7,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/project-costs-summary/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "tool-and-material-staging",
"version": "1.0.0",
"price_usd": 12,
"category": "Local Services / Field Operations",
"bundle_hash_short": "46db52eb08b8",
"bundle_kb": 34.7,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/roof-and-bathroom-staging/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "end-of-day-job-log",
"version": "1.0.0",
"price_usd": 12,
"category": "Local Services / Field Operations",
"bundle_hash_short": "c8fdb2f3db97",
"bundle_kb": 32.2,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/deck-rebuild-day-two/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "gift-card-and-credit-tracking",
"version": "1.0.0",
"price_usd": 12,
"category": "Local Services / Money",
"bundle_hash_short": "9a2b0eec87f0",
"bundle_kb": 29.1,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/gift-card-full-redemption/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "review-response-drafts",
"version": "1.0.0",
"price_usd": 12,
"category": "Local Services / Marketing",
"bundle_hash_short": "e7b9b270f8a4",
"bundle_kb": 31.7,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/hard-fence-repair-review/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "thumbnail-and-title-tests",
"version": "1.0.0",
"price_usd": 15,
"category": "Creator and Audience",
"bundle_hash_short": "7cb3a7c0a94b",
"bundle_kb": 38.9,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/three-dinners-from-one-chicken/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "podcast-guest-pipeline",
"version": "1.0.0",
"price_usd": 15,
"category": "Creator and Audience",
"bundle_hash_short": "0058a5e681ff",
"bundle_kb": 32.0,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/indie-maker-first-booking/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "new-agent-intake-checks",
"version": "1.0.0",
"price_usd": 15,
"category": "Agent Safety and Governance",
"bundle_hash_short": "629489d46970",
"bundle_kb": 41.5,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/inbox-triage-agent-for-a-trades-business/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "agent-fleet-roster",
"version": "1.0.0",
"price_usd": 12,
"category": "Agent Operations",
"bundle_hash_short": "0e3fab021022",
"bundle_kb": 39.6,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/first-roster-for-a-three-agent-studio/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "runbook-writing",
"version": "1.0.0",
"price_usd": 15,
"category": "Studio / Publishing",
"bundle_hash_short": "4ba69f557751",
"bundle_kb": 41.7,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/nightly-backup-failure/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "build-verification-checks",
"version": "1.0.0",
"price_usd": 12,
"category": "Studio / Shipping",
"bundle_hash_short": "e0332003a75e",
"bundle_kb": 39.9,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/mobile-app-blank-screen-catch/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "research-source-grading",
"version": "1.0.0",
"price_usd": 12,
"category": "Data, Files, and Research",
"bundle_hash_short": "594836028ac4",
"bundle_kb": 45.7,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/grading-a-vendor-performance-claim/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "file-metadata-cleanup",
"version": "1.0.0",
"price_usd": 12,
"category": "Data, Files, and Research",
"bundle_hash_short": "fcc26ba71f03",
"bundle_kb": 49.7,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/client-document-archive/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "first-conversation-scripts",
"version": "1.0.0",
"price_usd": 12,
"category": "Local Services / Sales",
"bundle_hash_short": "ea7805d70840",
"bundle_kb": 33.6,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/fence-repair-first-call/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "add-on-offers",
"version": "1.0.0",
"price_usd": 15,
"category": "Local Services / Sales",
"bundle_hash_short": "3d0cd3a99ffe",
"bundle_kb": 34.2,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/deck-repair-stain-and-seal/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "progress-payment-schedule",
"version": "1.0.0",
"price_usd": 15,
"category": "Local Services / Money",
"bundle_hash_short": "a68665ea8060",
"bundle_kb": 36.5,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/kitchen-remodel-three-stages/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "fuel-and-mileage-log",
"version": "1.0.0",
"price_usd": 12,
"category": "Local Services / Money",
"bundle_hash_short": "7d7b3d60f65d",
"bundle_kb": 33.7,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/fence-repair-two-jobs-one-van/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "site-safety-walk",
"version": "1.0.0",
"price_usd": 12,
"category": "Local Services / Field Operations",
"bundle_hash_short": "e45376abb654",
"bundle_kb": 38.8,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/brush-clearing-deck-teardown/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "subcontractor-payment-records",
"version": "1.0.0",
"price_usd": 15,
"category": "Local Services / Field Operations",
"bundle_hash_short": "d2a4535932a8",
"bundle_kb": 35.2,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/deck-framing-staged-payments/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "agent-decision-log",
"version": "1.0.0",
"price_usd": 15,
"category": "Agent Operations",
"bundle_hash_short": "c5e51c8cf103",
"bundle_kb": 47.1,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/a-pattern-review-that-changed-an-ordering-rule/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "agent-capacity-planning",
"version": "1.0.0",
"price_usd": 12,
"category": "Agent Operations",
"bundle_hash_short": "55781fd1f17e",
"bundle_kb": 47.2,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/four-agent-fleet-before-the-holiday-peak/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "secret-rotation-schedule",
"version": "1.0.0",
"price_usd": 15,
"category": "Agent Safety and Governance",
"bundle_hash_short": "4480723cf40b",
"bundle_kb": 41.0,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/catching-up-an-overdue-schedule/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "page-accessibility-checks",
"version": "1.0.0",
"price_usd": 12,
"category": "Studio / Publishing",
"bundle_hash_short": "131d4a167b36",
"bundle_kb": 44.7,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/event-listing-and-guide-pages-for-a-community-site/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "dependency-license-ledger",
"version": "1.0.0",
"price_usd": 12,
"category": "Studio / Shipping",
"bundle_hash_short": "5336fc2d18b8",
"bundle_kb": 41.6,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/first-ledger-for-a-shipping-app/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "archive-search-index",
"version": "1.0.0",
"price_usd": 12,
"category": "Data, Files, and Research",
"bundle_hash_short": "d401f5994c7e",
"bundle_kb": 43.6,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/first-index-for-a-contracts-archive/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "seasonal-booking-calendar",
"version": "1.0.0",
"price_usd": 12,
"category": "Local Services / Sales",
"bundle_hash_short": "6acb317e2698",
"bundle_kb": 36.0,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/exterior-washing-gutters-lights/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "materials-price-watch",
"version": "1.0.0",
"price_usd": 15,
"category": "Local Services / Money",
"bundle_hash_short": "45041066fa7f",
"bundle_kb": 34.2,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/fence-and-deck-may-check/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "crew-timesheet-records",
"version": "1.0.0",
"price_usd": 12,
"category": "Local Services / Field Operations",
"bundle_hash_short": "cbf0822e8877",
"bundle_kb": 38.3,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/three-person-crew-one-job-week/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "vehicle-and-trailer-checks",
"version": "1.0.0",
"price_usd": 12,
"category": "Local Services / Field Operations",
"bundle_hash_short": "74d2fd5195cf",
"bundle_kb": 34.6,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/clean-check-haul-to-job/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "data-entry-verification",
"version": "1.0.0",
"price_usd": 12,
"category": "Data, Files, and Research",
"bundle_hash_short": "22f34dd538bf",
"bundle_kb": 30.5,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/forty-eight-delivery-notes-stock-sheet/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "contact-list-hygiene",
"version": "1.0.0",
"price_usd": 12,
"category": "Data, Files, and Research",
"bundle_hash_short": "e4f711398257",
"bundle_kb": 39.8,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/small-contractor-list-cleanup/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "incident-postmortem-writing",
"version": "1.0.0",
"price_usd": 15,
"category": "Studio / Publishing",
"bundle_hash_short": "1be041173f08",
"bundle_kb": 55.1,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/checkout-timeout-postmortem.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "agent-workload-balancing",
"version": "1.0.0",
"price_usd": 12,
"category": "Agent Operations",
"bundle_hash_short": "e1ccd049fa31",
"bundle_kb": 42.6,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/overloaded-lookup-agent-relieved-by-two-moves/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "third-party-skill-watch",
"version": "1.0.0",
"price_usd": 12,
"category": "Agent Safety and Governance",
"bundle_hash_short": "53aac98f58ca",
"bundle_kb": 50.2,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/a-small-patch-reviewed-and-adopted.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "release-candidate-freeze",
"version": "1.0.0",
"price_usd": 12,
"category": "Studio / Shipping",
"bundle_hash_short": "11297c65d88e",
"bundle_kb": 49.3,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/freeze-with-one-admitted-and-one-held-out.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "visual-evidence-records",
"version": "1.0.0",
"price_usd": 12,
"category": "Data, Files, and Research",
"bundle_hash_short": "e06b4ea2893b",
"bundle_kb": 48.4,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/bug-report-checkout-failure/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "bulk-file-conversion",
"version": "1.0.0",
"price_usd": 12,
"category": "Data, Files, and Research",
"bundle_hash_short": "bf36b5f5c12b",
"bundle_kb": 46.6,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/course-handouts-doc-to-pdf/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "quote-expiry-and-renewal",
"version": "1.0.0",
"price_usd": 12,
"category": "Local Services / Sales",
"bundle_hash_short": "ca58f79adbed",
"bundle_kb": 34.7,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/fence-repair-renewal-after-a-rise/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "job-cancellation-handling",
"version": "1.0.0",
"price_usd": 15,
"category": "Local Services / Sales",
"bundle_hash_short": "b5b40bc6966c",
"bundle_kb": 33.1,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/fence-build-cancelled-after-materials/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "year-end-money-packet",
"version": "1.0.0",
"price_usd": 15,
"category": "Local Services / Money",
"bundle_hash_short": "b67eee30a76c",
"bundle_kb": 41.2,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/firewood-and-hauling-mixed-records-packet.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "dump-and-disposal-runs",
"version": "1.0.0",
"price_usd": 12,
"category": "Local Services / Field Operations",
"bundle_hash_short": "ee1fe0a9b7e2",
"bundle_kb": 33.6,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/shared-run-three-jobs/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "warranty-claims-packet",
"version": "1.0.0",
"price_usd": 15,
"category": "Local Services / Money",
"bundle_hash_short": "ec522ee9233a",
"bundle_kb": 40.1,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/circular-saw-motor-failure-claim.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "form-letter-library",
"version": "1.0.0",
"price_usd": 12,
"category": "Data, Files, and Research",
"bundle_hash_short": "804b8bb9d943",
"bundle_kb": 28.5,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/estimate-follow-up-and-balance-note/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "agent-handover-briefs",
"version": "1.0.0",
"price_usd": 15,
"category": "Agent Operations",
"bundle_hash_short": "e0131c759f94",
"bundle_kb": 50.8,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/a-weekend-cover-brief-for-a-repair-shop-inbox/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "approval-queue-management",
"version": "1.0.0",
"price_usd": 15,
"category": "Agent Safety and Governance",
"bundle_hash_short": "6187fec23b30",
"bundle_kb": 41.2,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/a-weekly-sweep-that-cleared-a-backlog/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "audit-trail-exports",
"version": "1.0.0",
"price_usd": 12,
"category": "Agent Safety and Governance",
"bundle_hash_short": "973190cb7b21",
"bundle_kb": 50.3,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/a-handover-packet-when-the-live-system-was-gone/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "release-checklist-runner",
"version": "1.0.0",
"price_usd": 12,
"category": "Studio / Shipping",
"bundle_hash_short": "c673ad41712d",
"bundle_kb": 39.1,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/a-release-held-at-the-announce-step/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "document-version-control",
"version": "1.0.0",
"price_usd": 12,
"category": "Data, Files, and Research",
"bundle_hash_short": "f4eb7d3a00e8",
"bundle_kb": 44.1,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/client-intake-pack-brought-under-control/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "meeting-agenda-builder",
"version": "1.0.0",
"price_usd": 12,
"category": "Agent Operations",
"bundle_hash_short": "a946f0afca31",
"bundle_kb": 38.7,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/launch-readiness-meeting-with-decisions-owed/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "inbox-zero-triage",
"version": "1.0.0",
"price_usd": 15,
"category": "Data, Files, and Research",
"bundle_hash_short": "b7dfd1541687",
"bundle_kb": 29.7,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/contractor-inbox-monday/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "recurring-task-scheduler",
"version": "1.0.0",
"price_usd": 12,
"category": "Agent Operations",
"bundle_hash_short": "f226a3d4f748",
"bundle_kb": 32.8,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/monthly-close-review/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "price-list-maintenance",
"version": "1.0.0",
"price_usd": 15,
"category": "Local Services / Money",
"bundle_hash_short": "edc593298c41",
"bundle_kb": 33.3,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/midyear-update-hauling-and-repairs/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "customer-offboarding-pack",
"version": "1.0.0",
"price_usd": 12,
"category": "Local Services / Sales",
"bundle_hash_short": "b0f042e1f2da",
"bundle_kb": 31.0,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/landscaping-season-close/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "tax-document-gathering",
"version": "1.0.0",
"price_usd": 15,
"category": "Local Services / Money",
"bundle_hash_short": "917ae9d0918f",
"bundle_kb": 45.2,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/full-year-lawn-and-cleanout/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "social-proof-collector",
"version": "1.0.0",
"price_usd": 12,
"category": "Creator and Audience",
"bundle_hash_short": "c93f4fdce327",
"bundle_kb": 33.8,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/creator-course-feedback/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "multi-step-job-planner",
"version": "1.0.0",
"price_usd": 15,
"category": "Agent Operations",
"bundle_hash_short": "d4880e28e207",
"bundle_kb": 51.2,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/help-article-portal-move-in-nine-steps/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "failure-mode-drills",
"version": "1.0.0",
"price_usd": 19,
"category": "Agent Safety and Governance",
"bundle_hash_short": "c6f58d60458e",
"bundle_kb": 52.5,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/bad-input-batch-at-intake/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "access-review-cadence",
"version": "1.0.0",
"price_usd": 15,
"category": "Agent Safety and Governance",
"bundle_hash_short": "521849d7e1d5",
"bundle_kb": 43.0,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/monthly-cycle-for-a-small-support-team/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "incident-severity-grading",
"version": "1.0.0",
"price_usd": 12,
"category": "Studio / Publishing",
"bundle_hash_short": "7272c2ecc266",
"bundle_kb": 37.8,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/morning-when-the-store-went-down-and-a-typo-waited/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "file-retention-exceptions",
"version": "1.0.0",
"price_usd": 12,
"category": "Data, Files, and Research",
"bundle_hash_short": "42dd2e95a69e",
"bundle_kb": 45.2,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/project-drafts-held-for-a-client-query/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "spreadsheet-formula-audit",
"version": "1.0.0",
"price_usd": 15,
"category": "Data, Files, and Research",
"bundle_hash_short": "f1a988efc474",
"bundle_kb": 57.2,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/monthly-budget-with-a-broken-total/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "quote-terms-library",
"version": "1.0.0",
"price_usd": 12,
"category": "Local Services / Sales",
"bundle_hash_short": "1142c0f077c4",
"bundle_kb": 35.0,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/kitchen-remodel-terms-caught-before-sending/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "tool-checkout-log",
"version": "1.0.0",
"price_usd": 12,
"category": "Local Services / Field Operations",
"bundle_hash_short": "f20332d8735c",
"bundle_kb": 31.7,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/fence-line-clearing-loppers-did-not-come-back/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "invoice-dispute-handling",
"version": "1.0.0",
"price_usd": 15,
"category": "Local Services / Money",
"bundle_hash_short": "6461c851645f",
"bundle_kb": 37.8,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/gutter-job-elbow-line-held-pipe-line-removed/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "lead-source-tracking",
"version": "1.0.0",
"price_usd": 12,
"category": "Local Services / Sales",
"bundle_hash_short": "32c12f100138",
"bundle_kb": 40.0,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/marketplace-heavy-month-for-a-gutter-operator/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "release-notes-drafts",
"version": "1.0.0",
"price_usd": 12,
"category": "Studio / Publishing",
"bundle_hash_short": "6837a0fe86b5",
"bundle_kb": 33.2,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/a-fix-only-release-with-nothing-new/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "audience-question-bank",
"version": "1.0.0",
"price_usd": 12,
"category": "Creator and Audience",
"bundle_hash_short": "60adaca8a91e",
"bundle_kb": 39.1,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/bakery-counter-questions-that-became-an-faq/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "agent-retrospective-notes",
"version": "1.0.0",
"price_usd": 12,
"category": "Agent Operations",
"bundle_hash_short": "6bd55bc70e1b",
"bundle_kb": 43.6,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/a-busy-month-that-changed-how-reminders-were-sent/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "prompt-version-registry",
"version": "1.0.0",
"price_usd": 15,
"category": "Agent Operations",
"bundle_hash_short": "88056c1dda87",
"bundle_kb": 52.2,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/bringing-an-old-prompt-back-after-a-bad-change/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "tool-permission-scoping",
"version": "1.0.0",
"price_usd": 15,
"category": "Agent Safety and Governance",
"bundle_hash_short": "28eb75083623",
"bundle_kb": 47.6,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/adding-a-payment-tool-with-a-tight-scope/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "canary-release-checks",
"version": "1.0.0",
"price_usd": 12,
"category": "Studio / Shipping",
"bundle_hash_short": "48e658dbe85f",
"bundle_kb": 48.8,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/a-clean-widening-that-reached-everyone/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "pdf-table-extraction",
"version": "1.0.0",
"price_usd": 15,
"category": "Data, Files, and Research",
"bundle_hash_short": "31bff007ad2e",
"bundle_kb": 46.9,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/a-rotated-statement-table-with-flagged-cells/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "naming-convention-setup",
"version": "1.0.0",
"price_usd": 12,
"category": "Data, Files, and Research",
"bundle_hash_short": "fab2535d62f5",
"bundle_kb": 53.2,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/designing-a-convention-for-a-new-research-study/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "newsletter-sponsor-slots",
"version": "1.0.0",
"price_usd": 15,
"category": "Creator and Audience",
"bundle_hash_short": "6f413a259cb9",
"bundle_kb": 33.5,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/autumn-launch-four-issues/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "community-mod-queue",
"version": "1.0.0",
"price_usd": 12,
"category": "Creator and Audience",
"bundle_hash_short": "3837f5ed4f28",
"bundle_kb": 32.5,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/creator-community-evening-queue/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "seasonal-crew-hiring-pack",
"version": "1.0.0",
"price_usd": 15,
"category": "Local Services / Field Operations",
"bundle_hash_short": "a02c1041495d",
"bundle_kb": 35.4,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/autumn-yard-crew-hiring-run/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "cash-handling-log",
"version": "1.0.0",
"price_usd": 12,
"category": "Local Services / Money",
"bundle_hash_short": "08dba643dac0",
"bundle_kb": 37.6,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/clean-week-reconciled-in-full/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "service-guarantee-tracking",
"version": "1.0.0",
"price_usd": 12,
"category": "Local Services / Reputation",
"bundle_hash_short": "74306d1fd531",
"bundle_kb": 34.0,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/kitchen-and-bath-ledger-opening/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "photo-backup-routine",
"version": "1.0.0",
"price_usd": 12,
"category": "Data, Files, and Research",
"bundle_hash_short": "39cd9ff1ad59",
"bundle_kb": 34.8,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/mixed-jobs-discrepancy-held/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "multi-agent-handoff-tests",
"version": "1.0.0",
"price_usd": 15,
"category": "Agent Operations",
"bundle_hash_short": "0bbe5992a436",
"bundle_kb": 53.9,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/eight-cases-for-a-draft-to-review-handoff/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "data-quality-spot-checks",
"version": "1.0.0",
"price_usd": 12,
"category": "Data, Files, and Research",
"bundle_hash_short": "2e61f3c36c07",
"bundle_kb": 43.8,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/monthly-supplier-records-check/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "secret-sharing-protocol",
"version": "1.0.0",
"price_usd": 15,
"category": "Agent Safety and Governance",
"bundle_hash_short": "e7f0b19a5628",
"bundle_kb": 39.9,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/giving-a-tool-one-afternoon-of-reach/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "uptime-incident-roster",
"version": "1.0.0",
"price_usd": 12,
"category": "Studio / Shipping",
"bundle_hash_short": "ee6e67578d51",
"bundle_kb": 42.2,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/four-person-team-fixes-night-handoffs/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "docs-link-checking",
"version": "1.0.0",
"price_usd": 12,
"category": "Studio / Publishing",
"bundle_hash_short": "ea3b0dcdfc94",
"bundle_kb": 38.8,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/quarterly-sweep-that-found-moved-guides/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "translation-glossary-keeping",
"version": "1.0.0",
"price_usd": 15,
"category": "Data, Files, and Research",
"bundle_hash_short": "ef5357eaaae0",
"bundle_kb": 55.6,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/cleaning-up-after-two-variants-spread/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "estimate-vs-actual-review",
"version": "1.0.0",
"price_usd": 15,
"category": "Local Services / Money",
"bundle_hash_short": "3105d7a14ad7",
"bundle_kb": 42.7,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/deck-repair-review-that-changed-the-scope/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "local-event-sponsorship-log",
"version": "1.0.0",
"price_usd": 12,
"category": "Local Services / Marketing",
"bundle_hash_short": "e701834d4c0c",
"bundle_kb": 38.3,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/county-fair-banner-and-booth/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "equipment-rental-scheduling",
"version": "1.0.0",
"price_usd": 12,
"category": "Local Services / Field Operations",
"bundle_hash_short": "ae1591744418",
"bundle_kb": 36.3,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/chipper-extension-that-held/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "winterization-checklist-runs",
"version": "1.0.0",
"price_usd": 12,
"category": "Local Services / Field Operations",
"bundle_hash_short": "185c037eb5cd",
"bundle_kb": 47.0,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/single-family-house-first-run/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "quote-photo-attachments",
"version": "1.0.0",
"price_usd": 12,
"category": "Local Services / Sales",
"bundle_hash_short": "db5ceebd51af",
"bundle_kb": 34.0,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/deck-repair-quote-with-labeled-photos/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "short-video-caption-checks",
"version": "1.0.0",
"price_usd": 12,
"category": "Creator and Audience",
"bundle_hash_short": "bf78a117d253",
"bundle_kb": 34.1,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/kitchen-tip-name-and-time-fixed/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "agent-fleet-cost-review",
"version": "1.0.0",
"price_usd": 15,
"category": "Agent Operations",
"bundle_hash_short": "f7b1878ab97d",
"bundle_kb": 41.2,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/first-review-for-a-four-agent-fleet/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "launch-day-runbook",
"version": "1.0.0",
"price_usd": 15,
"category": "Studio / Shipping",
"bundle_hash_short": "24d6ca1b1c0d",
"bundle_kb": 40.6,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/first-launch-for-a-small-tool/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "archive-integrity-checks",
"version": "1.0.0",
"price_usd": 12,
"category": "Data, Files, and Research",
"bundle_hash_short": "fd3d6456bdd3",
"bundle_kb": 41.5,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/first-check-for-a-small-office-archive/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "customer-win-back-notes",
"version": "1.0.0",
"price_usd": 12,
"category": "Local Services / Reputation",
"bundle_hash_short": "71fbf95e3d28",
"bundle_kb": 30.9,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/hauling-customer-after-a-long-gap/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "year-plan-budget-setup",
"version": "1.0.0",
"price_usd": 15,
"category": "Local Services / Money",
"bundle_hash_short": "57d2b56fa39b",
"bundle_kb": 41.2,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/first-year-plan-for-a-small-crew/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
},
{
"id": "skill-stack-planner",
"version": "1.0.0",
"price_usd": 19,
"category": "Agent Operations",
"bundle_hash_short": "586006291f01",
"bundle_kb": 41.4,
"composition": {"root": 1, "examples": 2, "reference": 4, "templates": 2},
"example_file": "examples/planning-a-monthly-report-workflow/WALKTHROUGH.md",
"source": {
"catalog_repository": "https://github.com/littlestjames82-sys/ghostbus-hub-skills",
"package_published": false
}
}
]
}
This catalog and the free Guardrails skill also live on GitHub: https://github.com/littlestjames82-sys/ghostbus-hub-skills
Install the free skill with the standard skills CLI: npx skills add littlestjames82-sys/ghostbus-hub-skills (tested Nov 4, 2026).
The Hub runs its own mail service, GhostMail. It is self-hosted, with no rented inboxes and no per-seat fees. The mail service itself passed 30 of 30 checks locally. It opens to real mail at launch, on the studio's own domain. Until then, everything on this page is local proof.
GhostPhone is built the same way as GhostMail: the studio runs its own switch, and agents hold extensions the way they hold mailboxes. The switch is our own SIP switch, written by the studio. Agents hold extensions: planner is 1001, hub-keeper is 1002, and buyer is 1003. Registration uses digest auth. Outbound calls are drafted and approved before anything dials. A real call carried voice media, and the media was verified byte for byte. Agents can text between extensions, and inbound calls are bridged to the bus. The service's own proof is 13 of 13. The local proof runs SIP and voice media over TCP, which real phone systems also support.
The building and proving happened locally. Launch is four taps from Ryan, then a short run of work from Muse. Nothing here is dressed up: this is the whole list.
Every change to the catalog reruns the Hub's whole regression gate against the real hub code. The latest run, Nov 6, 2026: ALL SUITES GREEN, 135 of 135 checks across 14 suites.
| Suite | Result | What it proves |
|---|---|---|
| Registry build | 200 skills | Rebuilds the catalog from the packages themselves; every listing's bundle hash is recomputed and must match. |
| Hub seed | 2028 files | Loads the full catalog into the hub store: 200 skills, 2028 files. |
| Proof 1: agent pull | 11 of 11 | An agent can find, pull, and verify a skill bundle end to end. |
| Proof 2: licenses | 12 of 12 | License keys issue and verify, and a held copy at an older version is flagged outdated. |
| Proof 3: hosted lifecycle | 10 of 10 | A hosted hub boots, serves the registry, and shuts down clean. |
| Proof 4: split vault | 12 of 12 | Paid bundles stay in the vault; only manifests and hashes are ever public. |
| Launch path | 11 of 11 | The public registry reads correctly with the hub key and the launch guardrails hold. |
| Per-buyer vaults and backups | 11 of 11 | Each buyer gets an isolated vault, and backups restore what they promise. |
| Seller approval flow | 9 of 9 | A third-party submission is scanned, approved, and listed without touching the launch catalog, and cleanup restores it exactly. |
| Seller phase 2 | 20 of 20 | Payouts, sale drafts, version updates, suspension, and delisting behave under the published terms. |
| Seller ops: sync, refunds, state backup | 10 of 10 | Catalog sync is a no-op on a fresh seed, pushes real changes, reverses payouts on refunds, and backs up operator state. |
| GhostMail: hub license email | 9 of 9 | License delivery drafts at an approval gate, sends only on approval, and refuses duplicates. |
| Round 2: commerce and operator tooling | 10 of 10 | The catalog export's hash verifies, and the operator tooling reports the hub truthfully. |
| Round 3: security and robustness | 10 of 10 | Hostile and malformed inputs are refused without corrupting hub state. |
The gate is one command, run locally by the studio on every catalog change. A red suite stops the change; tonight's three new skills went in only after it came back green.
Choose the domain the Hub and GhostMail will live on. Everything else keys off this choice, so it comes first.
Sign up for the small Hetzner server the Hub, the bus, and GhostMail will run on. The budget for this was already approved.
Open the Stripe account in the studio's name. Checkout opens at launch, after Stripe, and payments and seller payouts key off this account.
Decided on Nov 4, 2026: the Hub keeps 20% of each sale and the seller receives 80%. The seller terms are published as version v1 on the Terms page. This one is done.
No fine print voice. If the answer is not yet, this page says not yet.
Checkout is open now. Every paid skill has a Buy button that opens secure Stripe checkout for that exact skill and price. Cash App also works: pay inside Stripe checkout with Cash App Pay, or pay $RyanCotten2 directly in Cash App as described in the Cash App answer.
Because an agent that can spend money or speak for you should not do either on its own say so. Paid pulls and outbound messages wait for a human yes. That gate is the product working as designed, not a missing feature.
The Hub stores the hash of the key only. The key itself lives only in the buyer's email. If the Hub's records were ever read by the wrong eyes, there would be no usable keys in them.
The sale reverses in the payout ledger with the reason recorded, and the seller's still owed payout is adjusted. History is preserved, not erased. A payout that was already paid is settled manually.
No. GhostMail and GhostPhone are self-hosted: the studio runs its own mail server and its own SIP switch, and agents hold their own mailboxes and extensions. The one honest exception is real phone numbers and routes: nobody owns the phone network, so those rent carrier access through a trunk, by design.
Yes. GhostBus core is MIT and free to self host forever. Paid skills and hosted workspaces keep the lights on.
A permanent, personal, non-transferable licence to use it with your own agents, for your own work. The version you buy is the version licensed. Delivery is the bundle plus a license key by email, and the bundle hash on the listing is the record of what was delivered.
There is no bundle and no bundle checkout. Each skill is bought on its own, from its own Buy button at Stripe checkout. The Catalog paths show what a working set costs, summed plainly from the listed prices.
No. Every purchase is approved by the buyer or their human before it completes. That gate is the Hub's design, the same discipline the studio's own agent tools run under.
Compare the bundle's hash against the hash on its listing, and check the contents against the listing's What's inside file list. The full registry is verified on every run of the Hub's 135-check gate.
Email [email protected]. Support replies personally, with a target of one business day. Refunds are goodwill, with the license log as the record, and a confirmed problem delists the skill while it is reviewed.
The version you bought is the version licensed. Newer versions are new purchases unless the listing says otherwise. An update never replaces a live listing without a fresh scan and review, and the listing's vetting record is re-dated to the new scan.
The seller system is built and proven, and submissions open at launch. Seller terms are published on the Terms page, and the commission is 20 percent, so a seller keeps 80 percent of each sale.
Email [email protected]. Reports are read by a person, and a confirmed problem delists the skill while it is reviewed.
Every catalog change reruns a 14-suite regression gate against the real hub code: registry rebuild and hash verification, licensing, the split vault, seller flows and payouts, catalog sync, the license email gate, and hostile-input checks. The latest run, Nov 6, 2026, passed 135 of 135 checks. The full suite list is published on the Launch page.
Start with the free Agent Seatbelt Guardrails skill to see the gate work, then use the Catalog chooser to match a skill to your problem. Paid skills open at launch.
No. Badges are earned from actions on the Hub's own record and cannot be bought, gifted, or transferred.
Create an account when accounts open at launch, and the Hub awards badges automatically when your own record meets each rule, with the earned date taken from the record itself. Browsing, reading the catalog, and verifying a bundle with no account need no badge and no account. The Hub runs two badge systems. Member badges are earned by people from their own actions, and they are kept. Skill badges belong to listings, not people: each one is computed from the skill's own listing record, and if a listing loses the thing a badge stands for, the badge goes away. A seller cannot buy a skill badge or give one to their own skill.
Yes. Every paid skill can be bought today by paying $RyanCotten2 the exact listed price in Cash App, with the skill name and the email for delivery in the payment note. Ryan confirms each direct Cash App payment by hand before delivery by email under the Hub terms. Card checkout is open now for buyers who prefer it, and Cash App Pay is accepted inside Stripe checkout too.
Refunds are handled in goodwill, with the license log as the record, and there is no blanket guarantee. A pulled bundle cannot be un-pulled, so email [email protected] and a person will read the case in context.
No. A purchase grants a permanent, personal, non-transferable licence to use the skill with your own agents, for your own work. It does not cover resale, redistribution, or relisting the package or its contents.
Seller Terms, version v1.1. Published Nov 4, 2026. Version 1.1 adds the buyer licence. These terms govern selling on the Hub. Selling opens at launch.
Questions about these terms? Email [email protected].
Vetted is a record, not a badge of honor. Before a skill lists on the Hub, its package goes through the same checks every time, and the listing publishes what those checks found, with the date they ran. This page says exactly what those checks are, and what they are not.
Do not take our word for it: every listing publishes its file list and bundle hash, and the Help center shows you how to check one.
Every skill in the catalog passed the same gate before it listed:
The catalog reached 200 skills on November 10, 2026.
Every package is scanned two ways before it lists. The studio scanner reads every file in the package as text, and nothing else: it flags lines that match documented danger signals, such as secret harvesting, hidden sending of user files, instructions aimed at the reviewer, and disguised runnable code, and groups what it finds by severity. A second, independent tool, gitleaks, scans the same package for secrets that should never ship: keys, tokens, and credentials left inside files.
A scan signal is a lead for a reviewer, not a verdict. Pattern matching cannot tell a warning about a bad practice from an instance of one, so a person reads the findings from both scans before anything lists. Any high severity signal holds the package for human review, and a held package cannot be listed. Human approval is the only path from a submitted package to a live listing, for outside sellers and for the studio alike. The studio's own packages go through the same gate, including packages whose own text documents the signals the scanner looks for: those scan high because of what they describe, and they sit in the same review queue as everyone else's. First party packages written by the studio for this catalog list only with a clean result from the studio scanner and a clean gitleaks scan, and the Hub's own gate green for the wave they ship in.
Each listing carries the dated result of its scans, next to its bundle hash, in the open. The record names the scan date, the tools, and the outcome for that exact package. If a package changes, it is scanned again and the record is updated, because the hash below will have changed too.
Every listing publishes a bundle hash: a sha256 fingerprint of the whole package, computed the same way for every skill.
The method is published here and in the catalog's public repo, so anyone can recompute the hash of a delivered bundle and compare it with the listing. A match means the bytes delivered are exactly the bytes that were scanned and listed. A mismatch means they are not, and that is worth an email to [email protected].
Nothing about a listing is hidden except the paid package text itself. Every listing shows its full file list, a preview taken from the skill's own files, and a worked example, before anyone pays anything. The full catalog is also published in machine-readable form, with every skill's file list and bundle hash, so the record can be checked from outside the Hub.
The Hub scans packages as text. It does not execute them, install them, or run their instructions at any point in vetting. Vetting answers one question: what is in this package, and did it pass the scans on the date recorded. It is not a test run of the skill and it is not a promise about results in your setup. The listing's preview, file list, and worked example are there so you can judge fit before you buy.
The catalog itself is checked, not just the packages. The registry behind the Hub is hash-verified on every run of the Hub's own gate: 135 checks across 14 suites, covering the registry build, the catalog seed, licenses, the seller flow, and the mail path. The gate report is the studio's own, run on its own systems, and the current figure is published on this site and in the catalog repo.
The catalog's public record is the one that counts. If a delivered bundle's hash does not match its listing, or a listing's record looks wrong, write to [email protected] and it will be checked against the registry by hand.
Buying, delivery and license keys, refunds, verifying a bundle, selling on the Hub, and how to reach the studio, gathered in one place from the FAQ, the Terms, and the vetting record.
No help topic matches that filter. Clear the filter to see every topic, or email [email protected] with your question.
A human approves every purchase before it completes. An agent cannot complete a purchase on its own; the claim is refused until the buyer or their human approves it.
New to the Hub? Start free with Agent Seatbelt Guardrails, the Hub's one free skill, in the Start free section on the home page.
Delivery is the bundle plus a license key by email; the bundle hash on the listing is the record of what was delivered, and a buyer can verify a pulled bundle against it.
A permanent, personal, non-transferable licence to use it with your own agents, for your own work. The version you buy is the version licensed. Delivery is the bundle plus a license key by email, and the bundle hash on the listing is the record of what was delivered.
The Hub stores the hash of the key only. The key itself lives only in the buyer's email. If the Hub's records were ever read by the wrong eyes, there would be no usable keys in them.
The version you bought is the version licensed. Newer versions are new purchases unless the listing says otherwise. An update never replaces a live listing without a fresh scan and review, and the listing's vetting record is re-dated to the new scan.
No. A purchase grants a permanent, personal, non-transferable licence to use the skill with your own agents, for your own work. It does not cover resale, redistribution, or relisting the package or its contents.
Refunds are handled in goodwill, with the license log as the record, and there is no blanket guarantee. A pulled bundle cannot be un-pulled, so each request is read in context.
The sale reverses in the payout ledger with the reason recorded, and the seller's still owed payout is adjusted. History is preserved, not erased. A payout that was already paid is settled manually.
Every listing publishes a bundle hash: a sha256 fingerprint of the whole package, computed the same way for every skill. You can recompute it from the delivered files and compare it with the listing.
A match means the bytes delivered are exactly the bytes that were scanned and listed. A mismatch means they are not, and that is worth an email to [email protected]. The full method is published on the What vetted means page.
Pick the skill, paste the bundle hash from your delivery email, and press Check. The check runs on this page, against the hash published on the listing itself. Listings publish the short form of the hash, so a full hash from your email matches when it begins with the published short form.
Compare the bundle's hash against the hash on its listing, and check the contents against the listing's What's inside file list. The full registry is verified on every run of the Hub's 135-check gate.
The commission is 20 percent: the Hub keeps 20% of the sale price and the seller receives 80%. The rate used is recorded on every sale, so a later change never rewrites past sales.
Selling is curated. A seller registers with the operator first, giving their name, a contact email, and a payout method, and accepts the seller terms. Every package is scanned before review, and human approval is the only path to a listing.
Every Hub listing carries the same public elements, whether the skill is the studio's own or a seller's. A seller's listing will show:
Every submission goes through the same five steps published in How vetting works on the What vetted means page: a dual scan of the package, a human review of the findings, a dated record published on the listing, the full file list published with the listing, and a bundle hash a buyer can recompute from the delivered files. A skill that fails vetting is not listed.
Submissions are not open yet. To register interest today, email [email protected] with the subject "Selling on the Hub".
Email is the contact route for the Hub: [email protected]. Questions, refund requests, a problem with a listing, or a bundle whose hash does not match its listing all go to the same address, and a person reads them.