natebjones-projects/clean-my-ai-harness

Clean My AI Harness is a tool that inspects the hidden configuration files, custom instructions, and setup around your AI coding assistant, then shows you what is protecting your work and what is just adding friction, so you can safely clean it up. The first time you run it, nothing changes. It only produces a report, and you have to approve any changes yourself before the tool touches anything. There are two separate editions, one for Claude, covering Claude.ai and Claude Code, and one for Codex, covering OpenAI's Codex, and you download whichever zip file matches the AI tool you actually use. Each zip includes its own install instructions. For the Claude edition, you enable code execution and file creation, then upload the zip as a skill through the Customize menu. The Codex edition needs Python 3.10 or newer and gets copied into a specific skills folder in your home directory. Once installed, you open the project you want reviewed and ask the assistant to review its AI setup, starting in a read only mode that only shows what it would change. The output is a single visible report file that lists what shapes your AI's behavior, what protects your work, what is creating unnecessary drag, and what the cleaner recommends changing. You can approve specific numbered items in plain chat, such as approving items one and three while leaving item two alone, and the tool then writes a plain English summary of exactly what changed along with a way to roll the changes back. On safety, the README states that the cleaner treats every file it inspects as untrusted, never follows instructions hidden inside those files, and makes no changes until you explicitly approve each one. The license lets you use, modify, and share it freely, including in commercial work, with the one restriction being that you cannot resell the tool itself or a competing version of it.