caveman-compress

To compress markdown memory files, use the `/caveman-compress` command on your project files. The Skill processes the natural language text to reduce token consumption per session and automatically saves a human-readable backup of the original file.

Yes, compressing memory files requires both Python and the Claude API to function. The Skill relies on these dependencies to process markdown files and apply natural language compression for token reduction.

The best way to save input tokens on extensive project memory files is to compress their natural language text. This approach significantly reduces token consumption per session, resulting in lower costs over time while preserving a backup of the original data.

Yes, you can keep a human-readable version because the compression process automatically saves a backup of the original markdown file. This ensures you retain the full, uncompressed text for reference while using the compressed version for token efficiency.

You should compress markdown files for token efficiency when your projects have extensive memory files that increase load times and token consumption. Applying compression helps lower costs and improve session efficiency for large projects.