Three-Layer Safety for Autonomous Agents: Stopping the Infinite Loop

I watched an autonomous agent spend three hours and 40,000 tokens trying to close a GitHub issue that had an open dependency, only to fail because it kept hallucinating a force_close flag that didn't exist in the API. It didn't just fail; it entered a perfect infinite loop: it would call the tool, get a 400 error, interpret the error as a "temporary network glitch," and try again with the exact same payload. If you've built agents that actually touch production systems, you know this feeling. Pr