# donecheck

- **Event:** [Meta Global AI Developer Hackathon](https://cerebralvalley.ai/e/meta-ai-developer-hackathon)
- **When:** Oct 26, 12:00 PM – Nov 4, 9:00 PM (EST)
- **Where:** Online
- **Team:** [Edward Yi](https://cerebralvalley.ai/u/aiedwardyi)
- **GitHub:** https://github.com/aiedwardyi/donecheck
- **Demo video:** https://youtu.be/vJoe5LgsNZI
- **Gallery:** https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery
- **Page:** https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/4543

Coding agents often finish with "done, all tests pass" when they aren't done. donecheck reads an agent's completion report, pulls out each claim (files changed, the test command, other promises) and checks them against the repo. It compares claimed files with the real git diff, flags uncommitted work, warns about files changed but not mentioned, and runs the test command for real. It prints OK or FAIL per claim and exits non-zero if the agent overclaimed, so a script or pipeline can stop there. I run several coding agents in parallel every day, and checking their reports by hand is my bottleneck. This removes it.

## More from Meta Global AI Developer Hackathon

- [Team Fusion](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/4540)
- [Sonvane](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/4541)
- [Samosa AI](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/4542)
- [MetaVault OSS](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/4544)
- [Cumple&Cobra](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/4545)
- [Lumen AI](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/4546)

---

Markdown version of https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/4543. Site index for agents: https://cerebralvalley.ai/llms.txt · full text: https://cerebralvalley.ai/llms-full.txt