ai-ready
claudeKeep a repository ready for AI-agent work by enforcing what can be enforced and documenting only what cannot. audit writes a score-free gap report: which guidance docs exist (root and module AGENTS.md and the CLAUDE.md that imports it, docs/design decision records, a verification doc), which lint/typecheck/test/architecture checks exist and whether CI actually runs them, and every documented rule sorted into already enforced, cheap to enforce, not enforceable, or out of date. apply creates only what you approve: a short root AGENTS.md with a one-line CLAUDE.md that imports it (@AGENTS.md), module docs, decision-record pairs with union merge, an anti-pattern ledger, scripts/verify.sh with a git pre-push hook, a docs consistency check, and lint or architecture-test drafts for rules that only lived in prose. lessons turns corrections from the session and PR review into the same kinds of drafts. Release history lives in CHANGELOG.md.
3 skills · 1 agent