If your Claude Code auto-memory files are bloating with duplicates and misplaced rules, this walks you through a four-step cleanup: it finds entries already in CLAUDE.md or rules files, migrates project-specific rules out of memory into proper config locations, compresses what's left using tables and terse formatting, then validates the result. The interactive prompts let you review before each change, which matters since you're reorganizing the context Claude uses across sessions. Typical reduction is 30-50% on token count, which translates to faster loads and lower API costs. Works on the memory files in ~/.claude/projects, not your code.
npx skills add https://github.com/kochetkov-ma/claude-brewcode --skill memory-optimize