postmortem

Blameless Post-Mortem

Conduct a thorough, blameless analysis of an incident to understand what happened and prevent recurrence.

Instructions

Work through each section systematically. Focus on systems and processes, not individuals. The goal is learning, not blame.

Output Format

Executive Summary

Incident: [One-line description] Date/Time: [When it occurred] Duration: [How long] Severity: [Critical/High/Medium/Low]

Summary [2-3 sentence overview: what happened, impact, resolution]

---

Timeline

All times in [timezone].

Key Moments

• Trigger: [What initiated the incident]
• Detection: [How/when we knew something was wrong]
• Mitigation: [When bleeding was stopped]
• Resolution: [When fully resolved]

---

Impact

Customer Impact

• Users affected: [number]
• Duration of impact: [time]
• Symptoms experienced: [what users saw]

Business Impact

• Revenue impact: [if applicable]
• SLA/SLO breach: [yes/no, details]

---

Root Cause Analysis

Root Cause [The fundamental reason this happened—not the trigger, but why the trigger caused an incident]

Trigger [The immediate cause—what set things in motion]

Contributing Factors

Why Wasn't This Caught?

• Why didn't tests catch it?
• Why didn't monitoring alert earlier?

---

What Went Well

---

What Went Poorly

---

Action Items

Immediate (This week)

Short-term (This month)

---

Lessons Learned

1. [lesson]
2. [lesson]
3. [lesson]

---

Guidelines

• Blameless: "The deploy script didn't validate" not "Alice deployed without checking"
• Thorough timeline: Include everything, even embarrassing delays
• Specific actions: "Add memory alerting" not "Improve monitoring"
• Assigned owners: Unowned actions don't happen

$ARGUMENTS