Vibe Coding
Tip: Run /doctor and Treat CLAUDE.md as a Table of Contents, Not a Rule Dump
The built-in /doctor skill runs a nine-point read-only audit of a Claude Code setup: install health, duplicate installs, broken settings, colliding agent definitions, dead weight (skills/MCP servers/plugins with zero usage in your counters and transcripts), CLAUDE.md redundancy, lazy-loading opportunities, context budget, and permission settings. It reports first and asks before changing anything. In one live run it found 113 personal skills consuming roughly 10,000 tokens of context every session just to be listed, about a quarter of them never used. The recommended structure is a top-level CLAUDE.md that points to sub-files, with code review and verification moved into on-demand skills.
↳ Follow the thread