agentleFS
kunalsurikunalsuri/ai-fication-kit/.claude/skills/check-docs/SKILL.md

check-docs

Skill3 starsChanged 4 months agoApache-2.0View on GitHub

Kit-maintainer diagnostic (LOCAL-ONLY, not distributed) that audits the README and docs/ for staleness against the repo's real command/skill/CLI roster, stamped tree, and cross-doc consistency. Read-only; produces a severity-ranked findings report.

1 other project uses a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

7 rules. About 0 are common, 2 are distinctive, and 5 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈2

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 5

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Run the check-docs diagnostic.

    line 8
  2. Very specific:

    Use ls/Glob and grep for these — do not eyeball.

    line 47
  3. Very specific:

    When prose and code disagree, the code (templates/ + install.mjs) is the source of truth.

    line 129
  4. Distinctive:

    Does not auto-edit any file — it diagnoses and reports only.

    line 134
  5. Very specific:

    Does not audit the target-repo ai/ knowledge layer's file-path claims for drift against source — that is verify / drift / /check-drift.

    line 135
  6. Very specific:

    Does not judge ai/guide/ content quality — that is /post-cold-start-verification / /verify-ai-readiness.

    line 137
  7. Very specific:

    Does not get distributed to target repos — it is local to this kit (see .claude/LOCAL-ONLY-COMMANDS.md).

    line 139

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md
<!-- Copyright (c) 2026 Kunal Suri (CEA LIST). All rights reserved. -->

Run the check-docs diagnostic. Read-only. Produce a structured findings report; do NOT edit any file.

Local-only skill. This lives in .claude/skills/check-docs/ with no twin in templates/, so the installer never stamps it into target repos. It audits this kit's own README and docs — it is a maintainer tool, not a distributed workflow. See .claude/LOCAL-ONLY-COMMANDS.md. (Its own file, and any other local-only skill, is therefore expected to be absent from the distributed roster — do not flag that as drift.)

Why this exists

Docs drift silently: a feature adds a command, a skill, or a stamped file, and the prose roster, the What You Get tree, and the counts scattered across the docs quietly fall behind the code. verify/drift//check-drift guard the target-repo knowledge layer (ai/), not the kit's own public docs. This skill closes that gap.

Ground truth — derive these FIRST, from the code, before reading any prose

The distributed reality is defined by templates/, not by .claude/ (which also holds local-only extras like this skill and /implement-spec). Derive, deterministically:

  • Distributed workflows = the sub-directories of templates/claude/skills/ (the Claude surface — custom commands merged into skills). This is the canonical "what users get" roster. Cross-check parity with templates/github/prompts/, templates/agents/workflows/, and templates/cursor/rules/ (a workflow is "universal" only if present in all).
  • Distributed skills = the sub-directories of templates/claude/skills/ and templates/agents/skills/.
  • CLI commands = the members of the authoritative const COMMANDS = new Set([…]) in install.mjs (the CLI's own routing source, also used by test/release-check.mjs) — do not count command === "…" comparisons, which also appear for sub-flows. Mirrored in the docs/CLI-REFERENCE.md "Command index" table.
  • Stamped ai/ tree = the structure under templates/ai/ plus the files install.mjs writes directly (ai/repo-profile.json, ai/install-manifest.json, optionally ai/repo-indepth.json).
  • Local-only skills = present in .claude/skills/ but absent from templates/claude/skills/; the intended list is .claude/LOCAL-ONLY-COMMANDS.md.

Use ls/Glob and grep for these — do not eyeball. Every count you assert in a finding must trace to one of these sources.

Docs in scope

README.md, docs/README.md, docs/FAQ.md, docs/MULTI-TOOL-SETUP.md, docs/METHODOLOGY.md, docs/CLI-REFERENCE.md, docs/GETTING-STARTED.md, docs/reports/technical-report-draft.md, and the root CLAUDE.md / AGENTS.md.

Checks

Section D — Command & skill roster

IDCheckSeverity
D1Every prose count of workflows (e.g. README's "The Eleven Workflows" heading, "eleven workflows") equals the number of sub-directories in templates/claude/skills/❌
D2Every explicit workflow list/table (README roster table, MULTI-TOOL cross-tool table, technical-report §9.1) names exactly the sub-directories in templates/claude/skills/ — none missing, none removed❌
D3Every skill reference (README tree + highlights, FAQ, MULTI-TOOL-SETUP, technical-report §9.3) matches the sub-dirs of templates/claude/skills/ / templates/agents/skills/ — no skill named that isn't shipped, none shipped that isn't named❌
D4No local-only command (per .claude/LOCAL-ONLY-COMMANDS.md) is described anywhere as "stamped", "installed", or part of the distributed roster⚠️

Section E — CLI-command & stamped-tree drift

IDCheckSeverity
E1Every "all N commands" claim (e.g. docs/README.md's CLI-REFERENCE blurb) equals the size of the COMMANDS set in install.mjs❌
E2The docs/CLI-REFERENCE.md "Command index" table has one row per member of the COMMANDS set in install.mjs — none missing, none stale❌
E3The README "What You Get" tree lists every path templates/ai/ stamps and every directly-written ai/ file (spot-check: START-HERE.html, INDEX.md, lab/WORKLOG.md, lab/reviews/, guide/*, analysis/*)❌
E4The technical-report §8.1 directory tree agrees with the same stamped-tree ground truth⚠️

Section F — Cross-doc consistency

IDCheckSeverity
F1The command / skill / CLI counts agree with each other across all in-scope docs (a number stated in two places must match)❌
F2Version strings agree: package.json version vs the README citation version: vs CITATION.cff vs the newest CHANGELOG.md section⚠️
F3No doc references a file, command, or stack that no longer exists (e.g. a removed install.py / Python path, a renamed command)⚠️

Section G — Path-claim resolution

IDCheckSeverity
G1Every backtick-wrapped repo-relative path in the in-scope docs resolves on disk (same discipline verify applies to ai/, extended to README + docs/). Run node install.mjs verify . --strict first to clear the ai/ layer, then apply the path check to the docs above❌
G2Every relative markdown link [text](path) in README.md and docs/README.md resolves on disk⚠️

Output format

Emit the report directly to the user (do NOT write a file). Use this structure:

check-docs — findings
══════════════════════════════════════════════════

Ground truth (from templates/ + install.mjs)
────────────────────────────────────────────
Workflow commands: <N>  ·  Skills: <list>  ·  CLI commands: <M>
Local-only (not distributed): <list>

Roster (Section D)
──────────────────
✅/⚠️/❌  <ID> <check name>
           → <where it's wrong>: <one-line concrete fix>   ← omit for ✅

CLI & tree (Section E)
──────────────────────
✅/⚠️/❌  <ID> <check name>
           → <one-line concrete fix>

Cross-doc (Section F)
─────────────────────
✅/⚠️/❌  <ID> <check name>
           → <one-line concrete fix>

Path claims (Section G)
───────────────────────
✅/⚠️/❌  <ID> <check name>
           → <file:line>: <one-line concrete fix>

Summary
───────
X passed · Y warnings · Z errors
Next step: <single highest-priority action — one sentence>

Reporting discipline

  • Report every check, not just failures — passing checks build confidence.
  • For every ❌/⚠️, cite the exact file and, where possible, the line or heading that is wrong, and give a concrete one-line fix — never just "this is stale."
  • When prose and code disagree, the code (templates/ + install.mjs) is the source of truth; the fix always adjusts the prose, never the code.
  • The "Next step" names only the single highest-priority action.

What this command does NOT do

  • Does not auto-edit any file — it diagnoses and reports only.
  • Does not audit the target-repo ai/ knowledge layer's file-path claims for drift against source — that is verify / drift / /check-drift.
  • Does not judge ai/guide/ content quality — that is /post-cold-start-verification / /verify-ai-readiness.
  • Does not get distributed to target repos — it is local to this kit (see .claude/LOCAL-ONLY-COMMANDS.md).

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.