What is the purpose of a CLAUDE.md file?

CLAUDE.md is a project-specific guide that provides Claude with essential context regarding build commands, test patterns, and code style to improve its accuracy and autonomy.

Does it verify if my build commands actually work?

Yes, the skill includes automated validation scripts that test documented commands against your actual environment to ensure the AI doesn't receive hallucinated or outdated instructions.

When should I use the --thorough flag?

Use the --thorough flag when you want to evaluate 'Polish' tier criteria, which focus on fine-tuning the file for maximum token efficiency and readability beyond basic functionality.

Can this skill fix my files automatically?

Yes, by default the skill operates in --fix mode, which addresses failing checks by rewriting sections of your CLAUDE.md or creating rule files.

How does the scoring system work in this skill?

The skill uses a tiered rubric (Critical, Important, Polish) to evaluate the file and issues a verdict of PASS, NEEDS WORK, or FAIL based on how many checks are met.

Claude.md Auditor & Fixer

Name: Claude.md Auditor & Fixer
Author: smykla-skalski

bysmykla-skalski

•

Learning & Documentation

Audits and automatically repairs project context files to ensure optimal AI performance based on Anthropic best practices.

The review-claude-md skill provides a comprehensive auditing and repair workflow for CLAUDE.md and .claude/rules/ files. By evaluating project context against a tiered binary checklist of critical, important, and polish-level criteria, it ensures your AI agent has accurate instructions for builds, testing, and linting. The skill can either provide a 'score-only' verdict or automatically refactor documentation to meet high-performance standards, helping developers maintain efficient, low-token, and highly effective project contexts for Claude Code.

Key Features

01Intelligent codebase scanning to extract accurate project context

02Automated command verification to ensure build and test scripts are valid

03Detailed pass/fail reporting with verbose chain-of-thought reasoning

04Tiered checklist evaluation covering Critical, Important, and Polish criteria

052 GitHub stars

06Automatic repair of failing checks to align with official documentation standards

Use Cases

01Automating the maintenance of documentation as a project's build system evolves

02Optimizing project context files for better AI understanding and performance

03Standardizing AI instruction sets across multiple repositories in an organization

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add smykla-skalski/sai review-claude-md

For use in Claude.ai and ChatGPT

Key Features

01Intelligent codebase scanning to extract accurate project context

02Automated command verification to ensure build and test scripts are valid

03Detailed pass/fail reporting with verbose chain-of-thought reasoning

04Tiered checklist evaluation covering Critical, Important, and Polish criteria

052 GitHub stars

06Automatic repair of failing checks to align with official documentation standards

Use Cases

01Automating the maintenance of documentation as a project's build system evolves

02Optimizing project context files for better AI understanding and performance

03Standardizing AI instruction sets across multiple repositories in an organization

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add smykla-skalski/sai review-claude-md

For use in Claude.ai and ChatGPT