Review Claudemd

Spins up parallel Claude Sonnet subagents to analyze your conversation history and suggest improvements to both your global and project-specific CLAUDE.md files. It extracts recent conversations from ~/.claude/projects/, batches them by size, and has each subagent look for violated instructions, missing patterns, and outdated rules. The output groups findings into what needs reinforcement, what should be added locally versus globally, and what might be stale. Honestly most useful after you've noticed yourself repeating the same corrections across sessions. The batching logic is smart about conversation size, and the multi-agent approach means you're not waiting forever to process weeks of chat logs.

Install to Claude Code

npx -y skills add ykdojo/claude-code-tips --skill review-claudemd --agent claude-code