← ClaudeAtlas

healthlisted

Score documentation health across coverage, freshness, consistency, structure, discoverability, cross-linking, canonical integrity and metadata, and list the gaps the repository's own stack implies. Use for "how are my docs doing", documentation audits, or a periodic review.
ZeeshanSultan/DocGov · ★ 0 · AI & Automation · score 71
Install: claude install-skill ZeeshanSultan/DocGov
# Documentation health !`docgov health --json --compact 2>&1 | head -c 8000` ## How to report this 1. **The score, then the single highest-leverage fix.** Not eight bullet points of component scores the user has to rank themselves. Lead with: "91/100. The one thing worth fixing is X." 2. **Explain any component below 80 in terms of what is actually wrong** — the files, not the metric. "Freshness 67" means nothing; "four canonical documents have had no review while the code they describe changed 30 times" is a fact someone can act on. 3. **Coverage gaps are the most actionable part.** Each one means the repository's stack implies documentation that does not exist — Kubernetes manifests with no deployment guide, OAuth with no authentication model. Offer to create the top two or three with `/docgov:create`. 4. **Do not recommend raising the score.** Recommend fixing specific documents. A score that goes up because someone added frontmatter to archived notes is worse than a score that stays flat. ## Context for the numbers - **canonical integrity** below 100 is the serious one: it means the authority model itself has been violated, and every other number is built on sand. - **coverage** counts both what is classified and what the stack implies should exist. - **freshness** is semantic, not calendar-based — a correct document that has not changed in two years scores fine.