Agent Introspection Debugging
Community/About Agent Introspection Debugging
This skill provides a structured workflow for debugging failing AI agent runs. Instead of blindly retrying, the assistant learns to capture the failure state, diagnose the root cause, take a small contained recovery action, and finish with a readable self-debug report.
It is a workflow skill, not a hidden runtime, so it relies on the agent following the documented phases rather than on any special tooling.
AI-generated listing based on the project's public information. It may contain errors: check the original source.
/What Agent Introspection Debugging is for
My agent keeps hitting the maximum tool call limit, help me figure out why
The assistant is looping on the same command with no progress, diagnose it
The agent's output quality is degrading, check for context drift
A tool call failed with a connection error, capture the failure and suggest a recovery step
Write a self-debug report explaining what went wrong in this session