What problem does it solve? Project memory files like CLAUDE.md, todos, and preference notes are loaded into context on every session, and verbose prose inflates token costs repeatedly. This Skill rewrites those files into a terse caveman style while preserving all technical substance, cutting context tokens by roughly 46% on average. ## Core Features & Use Cases - Token-reducing compression: Rewrites natural language Markdown, text, reStructuredText, and TeX files into compact fragments while preserving code blocks, inline code, URLs, file paths, headings, and commands exactly. - Validated pipeline with retries: Detects file type, compresses via Claude, validates headings/code/URLs/paths against the original, applies targeted fixes for up to 2 retries, and restores the original on failure. - Safe backups and guardrails: Saves a human-readable FILE.original.md backup, refuses files over 500KB, skips code/config files, and blocks sensitive paths like credentials, secrets, or SSH keys. - Use Case: Run it on a 700-token CLAUDE.md preferences file to produce a ~285-token version that loads every session, with the original kept alongside for editing. ## Quick Start Ask the assistant to run /caveman-compress on your CLAUDE.md file to shrink it into caveman format with a backup saved automatically.
