caveman-compress

Compressing markdown project memory files minimizes input token consumption during AI sessions. This skill shrinks natural language prose into a concise format, reducing file size by an average of 46% while preserving code blocks and technical paths.

Compressing markdown documentation leaves code blocks, URLs, and file paths intact. The process strictly preserves technical instructions and structural integrity through a multi-stage validation and retry loop to prevent data loss.

You can recover your original markdown file because the compression process includes an automatic backup system. It creates .original.md files before modifying the active documentation to ensure your data is safely preserved.

You need both tiktoken and anthropic dependencies installed to run these markdown compression scripts. They provide the required environment for calculating token efficiency and executing the natural language compression process.

The best way to optimize CLAUDE.md files for AI session efficiency is to compress natural language into a concise format. This minimizes token overhead while keeping all functional instructions completely intact.