md-auditlisted

Audit human-facing markdown (README, CONTRIBUTING, SECURITY, docs/**, package READMEs) for claims the repo no longer satisfies — commands that fail, flags that don't exist, described behaviour the code dropped. Runs after scripts/audit-markdown.mjs has settled the mechanical findings. Use weekly from the docs-audit routine, or after a refactor that moved user-facing surfaces.
mattbutlerengineering/mattbutlerengineering · ★ 1 · Data & Documents · score 70

Install: claude install-skill mattbutlerengineering/mattbutlerengineering

# Markdown Audit (semantic pass) A broken link is a lookup. A false sentence is a judgement — and it is the more expensive of the two, because it reads as authoritative right up until someone runs the command and it fails. This skill audits the half of documentation rot that no script can settle: whether the prose is still **true**. ## Boundaries — read before starting Three passes exist and they do not overlap. Staying inside this one is what keeps the weekly routine cheap. | Pass | Owns | Cadence | | ---------------------------- | --------------------------------------------------------------------------- | ------- | | `scripts/audit-markdown.mjs` | Broken links, stale directory trees — anything decidable by `fs.existsSync` | weekly | | **this skill (`/md-audit`)** | Human-facing prose: claims about commands, flags, paths, ports, behaviour | weekly | | `/claude-md-improver` | Agent-instruction files: `CLAUDE.md`, `AGENTS.md`, `.claude/rules/*` | monthly | **Do not audit `CLAUDE.md`, `AGENTS.md`, or `.claude/rules/*`** — those belong to `/claude-md-improver` and duplicating it wastes a run and produces conflicting PRs. **Do not re-report broken links or tree entries** — the mechanical pass already fixed or reported them; if you see one, it is a known-unfixable the script deliberately left for a human. ## Scope In scope, in rough priority order: 1. `README