Agent Incident Postmortem

Write a blameless postmortem for this incident involving an AI agent.

Facts: {{incident_facts}}
Timeline notes: {{timeline_notes}}
Impact: {{impact}}

Structure: summary; impact (users, data, money, duration); timeline in UTC; how it was detected and how it SHOULD have been detected sooner; root cause and contributing factors (technical, process, prompt/model, policy) using the five-whys; what went well; what went poorly; action items each with an owner role, priority and due date, split into prevent / detect / mitigate; the specific eval cases and attack tests to add so this cannot silently recur; and communication that was or should be sent. Use neutral language about people. Mark any fact that is uncertain.