Agent Debugging Playbook - Fix Common Failures in 5 Minutes

Your agent broke and you do not know why. You check the logs (if there are any). You restart. You re-explain everything. Maybe it works, maybe it does not.

This playbook eliminates the guessing. It covers the 8 most common agent failure modes:

1. Context Overflow - agent gives wrong answers as sessions get long
2. Memory Loss - agent forgets everything between sessions
3. Stuck Loop - agent repeats the same action forever
4. Silent Stall - agent stops and waits instead of continuing
5. Instruction Drift - agent stops following rules over time
6. Progressive Degradation - agent gets worse week after week
7. Boundary Violation - agent does things you did not authorize
8. State Corruption - agent state files get corrupted after a crash

For each failure mode you get: specific symptoms to identify it, the root cause explained, step-by-step diagnostic checklist, a fix template you can drop into your workspace, and a verification test.

Built from real debugging sessions on a production autonomous agent. Every fix has been tested.

Works with any file-based agent. No code dependencies.