Skill Post-Mortem Facilitation Skill: универсальное решение для задач

--- name: "Post-Mortem Facilitation Skill" description: "Learn from failures without blame. Improve systems, not shame people." applyTo: "**/*incident*,**/*postmortem*,**/*retro*,**/*failure*,**/*outage*" --- # Post-Mortem Facilitation Skill > Learn from failures without blame. Improve systems, not shame people. ## Core Principle Every incident is a gift—an opportunity to make the system stronger. Blame prevents learning. ## The Blameless Philosophy | Blame Culture | Learning Culture | |---------------|------------------| | "Who messed up?" | "How did the system allow this?" | | "They should have known" | "Why wasn't it obvious?" | | "Follow the process!" | "Is the process followable?" | | "Don't let it happen again" | "How do we prevent this class of problem?" | **Key insight**: People did what made sense to them at the time, with the information they had. ## Post-Mortem Template ### 1. Summary (2-3 sentences) What happened, when, impact. ### 2. Timeline | Time (UTC) | Event | Actor/System | |------------|-------|--------------| | 14:00 | Deployment started | CI/CD | | 14:05 | Error rate increased | Monitoring | | 14:07 | On-call paged | PagerDuty | | ... | ... | ... | ### 3. Impact - **Duration**: How long? - **Users affected**: How many? - **Revenue impact**: If applicable - **Data loss**: Any? - **Reputation**: Customer communications? ### 4. Root Cause(s) Not "human error"—go deeper: - Why was the error possible? - What safeguards didn't catch it? - What systemic conditions contributed? ### 5. What Went Well - Detection time - Response coordination - Communication - Recovery speed ### 6. What Could Be Improved - Missing alerts - Documentation gaps - Process friction - Tool limitations ### 7. Action Items | Action | Owner | Due Date | Priority | |--------|-------|----------|----------| | Add validation | @alice | 2026-02-15 | P1 | | Improve runbook | @bob | 2026-02-10 | P2 | | ... | ... | ... | ... | **Rule**: Every action item has an owner and date. No orphan items. ## Facilitation Guide ### Before the Meeting 1. Gather timeline from logs, chat, alerts 2. Identify all participants (responders, stakeholders) 3. Set expectation: learning, not blame 4. Share draft timeline for review ### During the Meeting **Opening (5 min)** > "We're here to understand what happened and improve. This is blameless—we assume everyone acted reasonably with the info they had. Focus on systems and processes, not individuals." **Timeline Walk-through (20 min)** - Go chronologically - Ask: "What did you know at this point?" - Ask: "What options did you see?" - Fill in gaps in understanding **Root Cause Discussion (15 min)** - Use 5 Whys technique - Look for systemic issues - Avoid stopping at "human error" **Action Items (15 min)** - Prioritize by impact and effort - Assign owners IN the meeting - Set realistic due dates - Limit to 3-5 meaningful items (not 20 small ones) **Closing (5 min)** > "Thank you for the candid discussion. We'll share the write-up for review. Any final thoughts?" ### After the Meeting 1. Write up within 24 hours 2. Circulate for factual corrections 3. Publish to team/org 4. Track action items to completion 5. Review effectiveness in 30 days ## Anti-Patterns to Avoid | Anti-Pattern | Why It's Harmful | |--------------|------------------| | Naming individuals in root cause | Creates fear, hides future problems | | "They should have..." | Hindsight bias, doesn't fix systems | | No action items | Wasted learning opportunity | | Too many action items | Nothing gets done | | Action items without owners | Nothing gets done | | Never following up | Actions drift, cynicism grows | | Only for big incidents | Small incidents have big lessons | ## Questions That Unlock Learning - "What information would have helped at that moment?" - "What made sense to do at the time?" - "Where did our mental model differ from reality?" - "What surprised you?" - "What was harder than expected?" - "If this happens again, what would we do differently?" - "What's the smallest change that would have prevented this?" ## Severity Classification | Severity | Criteria | Post-Mortem Required? | |----------|----------|----------------------| | SEV1 | Customer-facing outage > 30min | Yes, within 48 hours | | SEV2 | Degraded service, workaround exists | Yes, within 1 week | | SEV3 | Internal impact, no customer effect | Recommended | | SEV4 | Near-miss, caught before impact | Optional but valuable | ## Connecting to Prevention Post-mortems feed into: - **Runbooks**: Better playbooks for next time - **Monitoring**: New alerts for early detection - **Testing**: New test cases for CI/CD - **Architecture**: Design changes to prevent recurrence - **Training**: Skills gaps identified ## The Phoenix Post-Mortem *A personal example from Alex's own evolution...* What we learned from the Phoenix incident: - Two sources of truth caused confusion - Testing in production (Master Alex) was risky - Kill switches need multiple layers - Document decisions as you make them These lessons became: ADR-006, RISKS.md, the 5-layer protection system. **Every failure makes the architecture stronger.** ## Synapses See [synapses.json](synapses.json) for connections.