What problem does it solve? Project memory files like CLAUDE.md, todos, and preference notes are loaded into context on every session start, so large files repeatedly consume input tokens. This Skill compresses those natural language files into a terse caveman format, cutting token usage by roughly 46% on average while preserving all technical substance. ## Core Features & Use Cases - Token-reducing compression: Rewrites prose by removing articles, filler, hedging, and redundant phrasing while keeping meaning intact. - Strict preservation rules: Code blocks, inline code, URLs, file paths, commands, headings, and technical terms are never modified. - Validation with targeted fixes: A local validator checks headings, code blocks, URLs, paths, and bullets; if errors are found, only the broken parts are patched via a targeted fix pass, with up to 2 retries and automatic restore of the original on failure. - Safe backup workflow: The original content is saved as FILE.original.md before overwriting, and existing backups are never overwritten. - Use Case: Run it on a 1000-token CLAUDE.md once, and every subsequent session loads the compressed version, saving hundreds of tokens per session automatically. ## Quick Start Ask the assistant to compress your CLAUDE.md memory file with the caveman-compress skill, for example by saying "compress memory file CLAUDE.md".
