review-agent-config
Skill3 starsChanged 4 months agoApache-2.0View on GitHub
Diagnostic gate that checks CLAUDE.md and AGENTS.md for completeness, consistency, and stale artifacts — run right after /cold-start and before /add-feature. Read-only; produces a findings report.
In this repository32 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- .github/copilot-instructions.mdCopilot instructions
- .cursor/rules/add-feature.mdcCursor rule
- .cursor/rules/adversarial-audit.mdcCursor rule
- .cursor/rules/ai-knowledge-layer.mdcCursor rule
- .cursor/rules/check-drift.mdcCursor rule
- .cursor/rules/cold-start.mdcCursor rule
- .cursor/rules/create-feature-catalog.mdcCursor rule
- .cursor/rules/fix-bug.mdcCursor rule
- .cursor/rules/perform-feature-add-simulation.mdcCursor rule
- .cursor/rules/post-cold-start-verification.mdcCursor rule
- .cursor/rules/review-agent-config.mdcCursor rule
- .cursor/rules/review-change.mdcCursor rule
- .cursor/rules/verify-ai-readiness.mdcCursor rule
- add-featureSkill
- cold-startSkill
- deep-testSkill
- fix-bugSkill
- review-changeSkill
- add-featureSkill
- adversarial-auditSkill
- check-docsSkill
- check-driftSkill
- cold-startSkill
- create-feature-catalogSkill
- fix-bugSkill
- implement-specSkill
- perform-feature-add-simulationSkill
- post-cold-start-verificationSkill
- review-changeSkill
- verify-ai-readinessSkill
The file
We split this skill into the rules it gives an agent
5 rules. About 0 are common, 3 are distinctive, and 2 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈3
Distinctive · estimate
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 8
Run the review-agent-config diagnostic.
- Distinctive: line 101
Does not auto-edit CLAUDE.md or AGENTS.md.
- Very specific: line 102
Does not check ai/guide/ content quality — that is /post-cold-start-verification.
- Very specific: line 103
Does not assess overall AI-layer maturity — that is /verify-ai-readiness.
- Distinctive: line 104
Does not run builds or tests.
≈ “Common” is an estimate: at least 1 in 5 of the 372,429 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.
Files
<!-- Copyright (c) 2026 Kunal Suri (CEA LIST). All rights reserved. -->Run the review-agent-config diagnostic. Read-only. Produce a structured findings report; do NOT edit any file.
Inputs — read these files (and only these)
CLAUDE.md(project root) — full readAGENTS.md(project root) — full readpackage.json(project root) — grep for"build"and"test"script keys onlypom.xml(project root, if present) — grep<build>section onlyai/guide/MODULE_MAP.md— check existence on disk; no deep read neededai/INDEX.md— check existence on disk; no deep read needed- Any
ai/guide/*.mdpaths referenced in the knowledge-map sections — existence check only via Glob
Checks
Section A — CLAUDE.md structure
| ID | Check | Severity |
| C1 | Contains the literal @AGENTS.md directive (not just a comment or a markdown link [AGENTS.md](AGENTS.md)) | ❌ |
| C2 | Has a "Hard rules" or equivalent section | ❌ |
| C3 | Has a "Where to look" or knowledge-map section with explicit pointers into ai/guide/ | ❌ |
| C4 | Has a token-discipline or subagent-delegation section | ⚠️ |
| C5 | No unfilled placeholders: regex <[a-zA-Z ]{3,}>, bare TODO, or the literal string fill in | ❌ |
| C6 | Build command is present and not a placeholder | ❌ |
| C7 | Test command is present and not a placeholder | ❌ |
| C8 | Test locations are filled in (not <fill in during cold start> or equivalent) | ❌ |
| C9 | No reference to CLAUDE_bkp_*.md files framed as active guidance (stale post-cold-start) | ⚠️ |
| C10 | No Claude-specific directives duplicated wholesale from AGENTS.md | ⚠️ |
Section B — AGENTS.md structure
| ID | Check | Severity |
| A1 | Contains no Claude Code-specific directives (@import, memory syntax, hook syntax) | ⚠️ |
| A2 | Hard rule present: frozen upstream / don't touch inherited code | ❌ |
| A3 | Hard rule present: test before done — with actual build and test commands | ❌ |
| A4 | Hard rule present: surgical diffs | ❌ |
| A5 | Hard rule present: provenance tagging ([inferred] / [verified]) | ❌ |
| A6 | Hard rule present: no phantom bugs / config churn | ❌ |
| A7 | Hard rule present: verify claims before declaring done | ❌ |
| A8 | Hard rule present: license-header matching for new files | ⚠️ |
| A9 | Has a knowledge-map section pointing to ai/guide/ | ❌ |
| A10 | No unfilled placeholders (same regex as C5) | ❌ |
| A11 | No reference to AGENTS_bkp_*.md files framed as active guidance | ⚠️ |
Section C — Cross-file consistency
| ID | Check | Severity |
| X1 | Build command in CLAUDE.md matches build command in AGENTS.md | ❌ |
| X2 | Test command in CLAUDE.md matches test command in AGENTS.md | ❌ |
| X3 | Build/test commands match what package.json scripts or pom.xml actually define | ⚠️ |
| X4 | At least one ai/guide/ path mentioned in any knowledge-map section resolves on disk | ❌ |
| X5 | No rule in CLAUDE.md directly contradicts a rule in AGENTS.md | ❌ |
| X6 | "Repo intelligence" / ai-layer description is not copy-pasted verbatim into both files | ⚠️ |
Output format
Emit the report directly to the user (do NOT write a file). Use this structure exactly:
review-agent-config — findings
══════════════════════════════════════════════════
CLAUDE.md
─────────
✅/⚠️/❌ <check ID> <check name>
→ <one-line concrete fix> ← omit this line for ✅
AGENTS.md
─────────
✅/⚠️/❌ <check ID> <check name>
→ <one-line concrete fix>
Cross-file
──────────
✅/⚠️/❌ <check ID> <check name>
→ <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 ❌ or ⚠️, always include a concrete one-line fix — never just "this is wrong." Name the exact line or section to change.
- The "Next step" must name only the single highest-priority action. Do not list all failures again.
- Severity governs the icon only; a ⚠️ warning is still reported and still gets a fix suggestion.
What this skill does NOT do
- Does not auto-edit CLAUDE.md or AGENTS.md.
- Does not check
ai/guide/content quality — that is/post-cold-start-verification. - Does not assess overall AI-layer maturity — that is
/verify-ai-readiness. - Does not run builds or tests.
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
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.

