Work through a production incident and write it up
From the first error trace to a blameless postmortem: stabilize, diagnose with ranked hypotheses, then document what happened and what to change.
Before you start
- Access to logs and the recent change history.
Steps
Stabilize first
Stop the bleeding (rollback, feature flag, scale) before investigating. Record timestamps as you go.
Check before moving on
- Customer impact has stopped or is contained.
Diagnose with hypotheses
Feed the stack trace and what changed into the diagnosis prompt, then run the quickest check for each hypothesis.
Uses: Diagnose a stack traceCheck before moving on
- At least one hypothesis is confirmed or ruled out with evidence.
Write the postmortem draft
Use the timeline you recorded to draft a blameless postmortem, then refine it with the team.
Check before moving on
- Every action item has a type; owners and due dates are assigned by humans.
Expected outcome
A resolved incident and a postmortem draft with action items.
Related resources
Prompt
Diagnose a stack traceTurns a stack trace plus a short description of what you were doing into ranked hypotheses, the first checks to run for each, and what extra evidence would settle it.
CuratedKitLocalUpdated 2 days ago
debuggingstack-tracetroubleshooting
About this catalog
Bundled snapshot · published Nov 15, 2026, 6:05 AM · Catalog revision 5bc079fe355af2214bc0
Showing the snapshot bundled with this app (no content source is configured). Newer content may exist at the source.