Inside the Package

What to look for

Every Skill in this package is one SKILL.md file. At the top of each sits a short description in YAML frontmatter — a sentence saying what the Skill does and when to use it. In the Skill format this package is written for, that description is the part a matching AI reads to decide whether this Skill is relevant to what you just asked. The instructions below it are what it follows once it has decided.

Here is the description from the use-case-triage Skill, as written:

Classify a proposed AI use case against your registry — approved, conditional, or not approved — and produce required conditions and next steps. Flags cross-plugin handoffs to privacy or product counsel. Use when user says “triage this use case”, “can we use AI for X”, “is this approved”, “what do we need to do to use AI for X”.

Notice how much of it is phrasing a person might actually type. That is deliberate, and it is specific to this Skill format rather than something every AI system does.

The five questions

Each one is about what a file is for. Open the reveal to see our pick and why.

Which file holds the steps the AI follows to triage a proposed use case?

Our pick: skills/use-case-triage/SKILL.md. A Skill is a single Markdown file, and this one carries both halves: the description at the top that says when the Skill applies, and the numbered instructions below it that say what to do — read the registry, clarify a vague request, check the red lines, classify, propose a registry update.

Where does the package keep what it has learned about your organization?

Our pick: CLAUDE.md. It is the practice profile: a setup interview written down once and then reused by every Skill in the package, so none of them has to ask again. It is the longest non-Skill file here, and it is shared context rather than instructions for any one task.

Which file supplies reference material a Skill reads when it needs it?

Our pick: references/currency-watch.md. A dated watch list of AI laws that move quickly, with a note about when the list itself should be rechecked. A Skill pulls it in when the question calls for it, which is the difference between supporting material and instructions.

Which file configures the outside services this package can reach?

Our pick: .mcp.json. Connector settings for this package’s environment. It is configuration rather than instructions, and it describes what this package was set up to reach — not something you are required to have, and not part of the Skill format itself.

Which file is housekeeping rather than anything the AI reads?

Our pick: .gitignore. Three lines telling version control what to skip. It ships inside the package because the package is also a repository, and it has nothing to do with what the AI is told.

What the AI actually uses

Working outward from a request, the order is roughly this. The description at the top of a SKILL.md is what gets that Skill picked. The instructions in the rest of that file are what it then follows. Shared context and reference files come in when the instructions say to read them.

The rest — the manifest, the README, the .gitignore — is there for people, for packaging and for version control. Those files are not simply handed to the model because they sit in the same folder.

The interactive version of this activity lets you search inside all sixteen files and read any of them in full, which needs JavaScript. The files themselves, and what each one is for, are listed below either way.

The sixteen files

Grouped the way the activity groups them: the Skills themselves, the shared material they draw on, and the files belonging to this package and its environment. The interactive version shows the full purpose note and the complete contents of each one.

Skills

Ten Skills, each a single SKILL.md under skills/.

• skills/ai-inventory/SKILL.md — track each AI system’s role and obligations
• skills/aia-generation/SKILL.md — run a structured AI impact assessment
• skills/cold-start-interview/SKILL.md — the setup interview that writes the practice profile
• skills/customize/SKILL.md — change one thing in the profile without redoing setup
• skills/matter-workspace/SKILL.md — create, list, switch and close matter workspaces
• skills/policy-monitor/SKILL.md — sweep saved assessments and keep the policy current
• skills/policy-starter/SKILL.md — draft a usage policy from published model policies
• skills/reg-gap-analysis/SKILL.md — diff a new regulation against the current posture
• skills/use-case-triage/SKILL.md — classify a proposed use case against the registry
• skills/vendor-ai-review/SKILL.md — review vendor AI terms against the profile

Shared context the Skills draw on

• CLAUDE.md — the practice profile, written once and reused by every Skill
• references/currency-watch.md — a dated watch list of fast-moving AI laws

Package and environment files

These belong to this package and the environment it was built for.

• .claude-plugin/plugin.json — the manifest: name, version, one-line description
• README.md — written for people: what it does and how to install it
• .mcp.json — connector settings for outside services this package can reach
• .gitignore — version-control housekeeping; the AI never reads it

If you want to build one of these rather than read one, AI Skills has the Library’s own downloadable Skills and the guides that come with them.