What problem does it solve? When a user corrects an AI agent's behavior, the lesson is often lost or applied as an over-fitted one-off rule. This Skill applies the immediate fix and captures the generalized lesson as a structured record, so guidance refinement happens later in bulk rather than reactively. ## Core Features & Use Cases - Immediate fix application: Applies concrete corrections to artifacts (code, docs, PR bodies) directly, asking for confirmation only when the change is ambiguous or risky. - Feedback classification: Distinguishes misapplied existing guidance (tagged mistake) from net-new desired behaviors, recording uncertainty when classification is unclear. - Structured record capture: Delegates to the capture-event skill to append a feedback-tagged record with summary, impact rating, and candidate refinement target to a project or default knowledge base. - Use Case: A user says "stop title-casing headings; the rule is sentence case." The agent fixes the heading, then records the error-correction pair tagged feedback,mistake for a later bulk guidance-refinement pass. ## Quick Start Tell the agent to use capture-feedback to apply your correction and record the lesson for future guidance refinement.
