SuperAPI AI Directory

audit-agent-docs

Coding & development

View repository

Updated 10-16 06:18

About this skill

Audit Agent Docs

Audit whether existing project docs give an agent enough facts to start work without guessing.

Do not edit by default. Report gaps first.

Inspect

Look for:

  • root agent instructions: AGENTS.md, CLAUDE.md, or equivalent
  • domain language: CONTEXT.md or CONTEXT-MAP.md
  • durable decisions: docs/adr/
  • agent docs: docs/agents/ or project equivalent
  • build, test, owner, security, and tool policy facts
  • package scripts, CI config, test config, and obvious entrypoints

If files are absent, inspect the repo directly before calling it a gap.

Audit Questions

The docs should let an agent answer:

  • Where should I start reading?
  • How do I install, build, run, lint, typecheck, and test?
  • Which modules are likely in scope?
  • Which paths or actions need approval?
  • Who owns decisions or review?
  • Which known risks should not be "fixed" accidentally?
  • Which tools, connectors, skills, hooks, and side effects are allowed?

For a deeper checklist, use readiness-questions.md.

Output

Return:

  • ready, ready_with_gaps, or not_ready
  • highest-risk gaps first
  • evidence: file path or command result
  • concrete fix for each gap
  • docs that are redundant or should become skills

Do not demand placeholder docs. If a fact is not needed for safe work, say so.