Agent Self-Evaluation helps an AI agent pause after completing a complex task to rate its own output against a structured five-axis rubric covering accuracy, completeness, clarity, actionability, and conciseness. It serves as a deliberate reflection step that catches omissions, flags overconfidence, and surfaces areas for improvement before the user has to review them. Reach for this skill when you need a rigorous, evidence-based quality check on non-trivial tasks like multi-file coding, debugging sessions, or written analysis.