agent-introspection

To run a post-mortem after a code error, you identify the root cause and record actionable rules into a LEARNINGS log to prevent repeats. This structured recovery improves future task completion and overall quality.

Clarifying questions are concise pre-flight prompts used to verify missing context before you start coding or making decisions. They are needed to ensure correctness and avoid avoidable mistakes early in the agent workflow.

You set up a self-audit checklist by defining review criteria for correctness, safety, cleanliness, and compatibility. Running this checklist before sending outputs catches issues early and enforces post-output guardrails during task completion.

Yes, structured weekly introspection improves agent workflow quality by answering curated prompts that uncover automation opportunities, drifting assumptions, and memory gaps. This iterative learning loop turns repeated patterns into future guardrails.

The best way to prevent repeated mistakes is establishing an iterative learning loop that records explicit root-cause captures into LEARNINGS. Turning these repeated patterns into proactive guardrails ensures continuous quality improvement.