I developed this because I kept typing the same requests to Claude Code/codex every day, worded a little differently each time ("did you send this email?", "how many request you got so far?").
This small Python script reads your local Claude Code, Codex and Cursor history for any project you want. One of my projects has no exact sentence repeated more than 14 times, but the top ask, grouped by intent, had been typed 230 times.
It only reads history still on your machine, and matching is per directory.
I developed this because I kept typing the same requests to Claude Code/codex every day, worded a little differently each time ("did you send this email?", "how many request you got so far?"). This small Python script reads your local Claude Code, Codex and Cursor history for any project you want. One of my projects has no exact sentence repeated more than 14 times, but the top ask, grouped by intent, had been typed 230 times.
It only reads history still on your machine, and matching is per directory.
Fun idea. How has your experience been so far? Any hiccups?
How to configure each agent to use agent skill slash commands automatically?