Use when an AI agent needs calibrated confidence, uncertainty-aware answers, evidence checks, or a pre-completion verification checklist. Helps prevent overconfident claims like "fixed", "done", "safe", or "definitely correct" unless supported by evidence.
