A blameless postmortem asks why the system let an agent take a harmful action, not why the agent misbehaved. Record the missing control, the detection delay and the permission that should not have existed, then turn each into a tracked fix. Public agent incidents are good practice material.
Read in this order
- BookPostmortem Culture: Learning from Failure (Google SRE Book)Start here: the principles of blameless postmortems from SRE practice.
- ArticleIncident 1469: PocketOS production database reportedly deleted by an AI coding agent (AI Incident Database)A 2026 data-loss incident to practice a blameless post-mortem on.
- OfficialPagerDuty Postmortem DocumentationA postmortem template and guide you can adopt as is.