agent-introspection-debugging

Debug an AI agent that repeatedly fails by capturing the failure state, diagnosing root causes, and applying contained recovery. This structured self-debugging prevents token waste and stops tool looping before escalation.

Structured self-debugging is a four-phase workflow for AI agent failures: Failure Capture, Root-Cause Diagnosis, Contained Recovery, and Introspection Report. It records evidence and ensures traceability for humans or future agents.

Diagnose root causes for an agent drifting from task objectives by capturing the failure state, mapping errors to known patterns, and surfacing actionable hypotheses. This structured diagnosis guides subsequent contained recovery actions.

Recover a failing AI agent without escalation by applying contained recovery, which uses small, reversible actions to regain progress. This approach safely addresses the diagnosed root causes while preventing further token waste.

Generate an introspection report for AI agent failures after completing failure capture, root-cause diagnosis, and contained recovery. It produces a structured record of evidence and traceability for humans or future agents.

Structured agent debugging requires no external dependencies or components to operate. It relies entirely on a four-phase workflow using a structured recovery checklist and templates to record evidence and ensure traceability.