Find why an agent produced a bad result, strengthen the code, repository, verification, tooling, or workflow that enabled it, then retry the original task. Use when the user asks to learn from a mistake, prevent recurrence, or postmortem an agent failure.
