compress

Compressing natural language memory files into caveman format reduces input tokens by overwriting the original file with a condensed version. The process preserves code blocks, URLs, and file paths, and saves a human-readable backup as FILE.original.md.

Caveman format compression is a technique that condenses natural language memory files like CLAUDE.md, todos, and preferences to save tokens. It preserves essential structural elements like code blocks, URLs, and file paths during the compression process.

Yes, compressing memory files saves a human-readable backup as FILE.original.md before overwriting the original. This backup ensures you can recover the full natural language content if the compressed caveman version needs rollback.

Run the compress command on a target file to generate a compressed caveman version. The tool automatically overwrites the original file with the condensed content and creates a backup named filename.original.md for recovery.

Yes, compressing markdown memory files into caveman format explicitly preserves code blocks, URLs, and file paths. Only the surrounding natural language text is condensed to achieve token savings.

You should avoid compressing memory files when full human readability is required for active editing or collaboration. The caveman format intentionally condenses natural language, which may make direct reading difficult without restoring the FILE.original.md backup.