Pluviobyte/jiepi-skill — reverse-engineered prompt

Reverse engineered prompt

Build me a local skill called Jiepi that helps me audit which AI agent skills I actually use on my machine. I want it to work as a skill inside Codex and Claude Code, but also be runnable as a normal Python script. It should scan the local skill folders and history data for Codex, Claude Code, and Cursor, then produce a simple report showing each skill, where it came from, how many confirmed times it was called, how many of those were in the last 30 days, how many sessions used it, and when it was last used. Give each one a cleanup suggestion like keep, review, or cleanup candidate. Please merge duplicate cache versions from the same plugin so counts are not inflated, and optionally scan project level skills under Desktop and Documents. Output should support Markdown and JSON, and let me change the recent usage window or save the report to a file. Use a conservative counting rule based on explicit SKILL.md reads.

Want more depth? Deep Reverse