I found a really cool tool that turns GitHub repos into missions for AI agents lol, an agent tries it for you - run, review, explain, roast it 😭

I’ve been experimenting with a tool that turns GitHub repos into little missions for AI agents. You paste a repo, choose a mode like run/review/roast, and it spins up an agent in a sandbox, records the desktop/terminal/browser, then returns a narrated video of the entire session. I tried a “brutally roast this repo” prompt on a random repo and the result is basically a DX review driven only by the actual agent run: install attempts, misconfigurations, dead ends, the moment it finally works, etc. It’s an interesting way to use agents as first‑pass repo evaluators with concrete replayable evidence. If you’re into agent workflows and tooling, I've pinned this tool in the comments :))