nadimtuhin/claude-token-optimizer — reverse-engineered prompt
Reverse engineered prompt
Build me a small command line tool for people who use Claude Code and keep wasting context on old notes and docs. I want someone to run one simple setup command inside any project and have it create a clean Claude documentation setup with only the important startup files loaded, plus folders for completed tasks, old sessions, archived docs, and topic based learnings that Claude can read only when asked.
It should be able to measure how many tokens are currently being loaded, initialize the optimized structure, detect common project types like Next, Express, Django, Rails, Laravel, Go, Spring Boot, Svelte, and others, and add useful starter notes about common mistakes for each one. Also add maintenance commands to audit the setup, compress the main Claude file, prune stale sections safely by archiving them, show token savings, and watch token usage live.
Please make it installable as a normal JavaScript CLI, include examples, tests, clear docs, and safe defaults so it never destroys a user’s custom notes.
Want more depth? Deep Reverse