shipping-artifacts is the reference the other ai-shipping skills share. It states what each system document must contain, what a reviewer uses it for, and where it lives, so every later audit has written intent to compare the code against.
Its Locations and names rule puts the host repo’s own docs folder, file names and date conventions first and uses its defaults (a documentation folder, tests.md, reports/) only where the repo has none.
What You Get
- Required contents for architecture.md, flows.md, permissions.md and variables.md, including a resource by operation by role matrix and a variables table with scope, rotation and risk
- A three-section shape for tests.md that separates existing coverage, proposed tests and gaps so the map cannot read falsely green
- Conditional document specs for emails, scheduled jobs, SEO previews and embedded agents, including each agent’s tool surface and guardrails
- A lookup procedure for the repo’s existing docs, coverage map and reports location, and a case-insensitive file-name collision rule
Common Use Cases
- Deciding which documents an AI-built app needs before an audit
- Matching a repo’s existing docs folder and updating its equivalents
- Giving a reviewer who has never seen the code enough to answer who may do what
- Documenting an app that embeds agents or webhooks
What Makes It Unique
It is a shared specification with no workflow of its own: document-app, derive-tests and the audits all point to it. It insists on describing the current state with short, table-heavy docs, records absent capabilities and never invents an empty document, and it guards against creating a file whose name differs from an existing one only in letter case.
Under The Hood
A single markdown file with a short receipts file beside it. It carries no scripts and runs no commands. The agent that applies it looks for agent instruction files, docs folders and existing equivalents in the repository, then reads and writes markdown documents there. It calls no web services.