AgentBugs
Structured evidence for what AI agents fail to do, what reproduces, and what actually works around it.
MVP shell for local development and review.
Focused on three actions only: search a failure, report a failure, and confirm or work around a failure.