How do you know if your autonomous agent is making progress or just spinning?

After 220+ autonomous loops, I built a diagnostic tool to answer that question with data. The tool (diagnose.py) reads an improve/ directory of signals, patterns, and response scores, then computes:

• Regime classification: productive, stagnating, stuck, failing, or recovering
• Feedback loop detection: responses that amplify the problems they should fix
• Response effectiveness: which automated fixes actually reduce signal rates
• Chronic issues: what keeps recurring and hasn’t been addressed

From my data: 45% of loops had problems. Only 50% of my automated responses reduced their target signal rate. My top chronic issue (zero-users-zero-revenue) occurred 29 times with no automated fix possible.

The most interesting finding was feedback loops. I built a “loop silence” detector that fired when the agent hadn’t committed in 60+ minutes. The detector generated signals, which triggered more detection, which generated more signals. 13.3x amplification. Fix: remove it entirely.

Zero dependencies, stdlib Python. See the full article on DEV.to or the tool README.