skills.rest
Yoodaddy0311
Yoodaddy0311COMMUNITY
Explore all skills by this author →
GitHub

claude-md-auditor

Audit CLAUDE.md memory files against a 6-criteria rubric and propose approval-gated diffs.

3|2|Updated Feb 13, 2026
Search Tags:#documentation#claude code#audit#claude.md#diff review#scoring rubric#memory quality
One-click install
npx skills add https://github.com/Yoodaddy0311/artibot --skill claude-md-auditor-yoodaddy0311
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: claude-md-auditor
Source: https://github.com/Yoodaddy0311/artibot/tree/main/plugins/artibot/skills/claude-md-auditor
Command: npx skills add https://github.com/Yoodaddy0311/artibot --skill claude-md-auditor-yoodaddy0311

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? CLAUDE.md memory files drift out of date, bloat with obvious content, and lose usefulness over time, but teams have no systematic way to measure or fix their quality. ## Core Features & Use Cases - Rubric-Based Scoring: Scores every CLAUDE.md file against 6 criteria (commands, architecture, patterns, conciseness, currency, actionability) on a 100-point scale with A–F grades. - Approval-Gated Fixes: Proposes minimal diffs (≤40 lines) for low-scoring files and only applies Edits after explicit user approval. - Metrics Caching: Writes structured results to an audit-cache.json file for downstream GRPO learning pipelines. - Use Case: After a learning cycle updates project memory, run the audit to verify no CLAUDE.md file exceeded the 200-line cap or went stale beyond 90 days. ## Quick Start Ask the assistant to audit the CLAUDE.md files in this repository and score their memory quality.

Frequently Asked Questions about claude-md-auditor

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I audit CLAUDE.md files for quality?▼

Run the audit to scan all CLAUDE.md, CLAUDE.local.md, and package-level memory files via Glob patterns. Each file is scored against 6 criteria totaling 100 points and graded A through F, with diffs proposed for files graded C or lower.

What criteria define a good CLAUDE.md file?▼

The rubric scores commands and workflows (20 pts), architecture clarity (20 pts), non-obvious patterns (15 pts), conciseness under 200 lines (15 pts), currency within 90 days (15 pts), and actionability with explicit triggers (15 pts).

Does the audit modify CLAUDE.md files automatically?▼

No. The audit only proposes minimal diffs of 40 lines or fewer and requires an explicit yes or select response before any Edit is applied. Without approval it defaults to skipping all changes.

When should I not run a CLAUDE.md audit?▼

Skip it when the user already specified a single concrete edit to a CLAUDE.md file, or when no CLAUDE.md file exists in the repository—in that case bootstrap one first with a learning command instead.

Why does the audit use Glob instead of find?▼

Glob is used because find is rejected by the pre-bash guard on Windows and is not portable across platforms. Discovery scans prioritized patterns including root, subdirectory, local override, global user, and plugin memory files.

Related Skills

Explore More Software Engineering Skills→
caseproofcaseproof

AI Memory System Skill

Manage persistent markdown memory files across AI Friday sessions.

Official
Basic
adelproadelpro
2

compound-engineering

Review past sessions, extract learnings, and update memory files.

Community
Advanced
bactopiabactopia
521

review-docs

Validate .claude/docs reference documentation for staleness and drift with bactopia-docs --validate.

Official
Intermediate
WizarckWizarck

claude-md-improver

Audit CLAUDE.md files and propose targeted improvements after approval.

Community
Advanced