borghei/Claude-Skills / project-management/execution/post-mortem
post-mortem
VerifiedRun a blameless review with a timeline, contributing causes and owned action items. Use when an incident, outage, missed launch or failed experiment needs a written review.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 10.3 KB |
| assets/action_item_tracker.md | markdown | 3.6 KB |
| assets/incident_timeline_worksheet.md | markdown | 3.6 KB |
| assets/post_mortem_template.md | markdown | 4.3 KB |
| assets/what_went_well_prompts.md | markdown | 6.1 KB |
| examples/payment-processing-outage.md | markdown | 14.8 KB |
| references/5-whys-vs-causal-tree.md | markdown | 12.4 KB |
| references/blameless-culture-guide.md | markdown | 10.3 KB |
| references/post-mortem-process.md | markdown | 6.5 KB |
| references/red-flags.md | markdown | 19.3 KB |
assets/what_went_well_prompts.md
View on GitHub# "What Went Well" Prompts
The "What Went Well" section is the most-often-skipped section of post-mortems and one of the highest-leverage. It:
- Surfaces practices, tooling, and decisions worth replicating across the team.
- Anchors the document in learning rather than blame avoidance.
- Builds psychological safety. A post-mortem that only catalogs failures becomes a punishment artifact even with the best intentions.
- Identifies what the team already does that the next incident should preserve.
Aim for a **minimum of 3 items**, ideally 5-7. If the section is genuinely empty, that itself is a finding — the team should investigate why nothing in the response went well.
## Prompts by phase
### Detection
- What signal first told us something was wrong? Was that signal designed to fire for this class of failure?
- How fast was detection compared to past incidents in this class?
- Did anyone notice through a non-paged channel (chat report, customer support, manual dashboard check)? That informal detection might be worth formalizing.
- Did a dashboard, an alert, or a log line make the cause obvious within seconds of looking?
### Communication
- Was the incident chat opened promptly? Was the right channel used?
- Did the team adopt clear roles (incident commander, scribe, communications) without confusion?
- Did external communication (status page, customer email, exec brief) go out at the right time and with the right tone?
- Was anyone notably good at translating between engineering detail and customer/exec language?
### Decision-making
- Was a critical decision made with limited information that, in hindsight, was the right call? That tells you something about the team's judgment process.
- Did the team avoid a tempting-but-wrong path (e.g. resisted "let's just restart everything" when the right move was to investigate first)?
- Was a reversible decision made quickly and an irreversible decision deliberately? That is decision hygiene worth naming.
### Tooling
- What tool, dashboard, or runbook step saved time?
- What recent investment in tooling paid off during this incident? (Especially if the investment was contested at the time.)
- What automation worked as designed?
### Mitigation
- What was the time-to-mitigation? Compare to past incidents.
- Did the rollback or feature-flag-disable work cleanly?
- Did failovers behave as designed?
### Containment
- Was customer impact smaller than it could have been? Why?
- Was the blast radius reduced by a recent architectural change (cell isolation, circuit breaker, rate limit)?
### Team behavior
- Did anyone bring outside expertise quickly (calling another team, an SRE on a non-rotational page)?
- Did anyone explicitly slow down to think when the team was in panic mode?
- Did a junior engineer ask a clarifying question that unlocked a path?
- Did the team avoid heroics — using documented process rather than improvisation under pressure?
### Recovery
- Was the customer-impact end clearly verified rather than assumed?
- Was the post-incident verification thorough?
- Did the team return to BAU work cleanly, or did the incident leave on-call exhausted? (Recovery wellbeing is part of post-incident health.)
## Anti-patterns when writing this section
- **Backhanded compliments.** "What went well: the team eventually figured it out." This is what-went-wrong dressed up. Remove.
- **Generic statements.** "Communication was good." Be specific: "The incident commander posted status updates every 10 minutes in #incident-payments, which kept exec stakeholders from interrupting the response work."
- **Listing the absence of bad things.** "Nobody panicked." Reframe to the positive presence: "The team paused for 90 seconds at minute 4 to confirm the leading hypothesis before acting."
- **Only naming tools, never people-system practices.** Tools matter, but so do practices. Both belong.
## Examples of good "what went well" entries
Drawn from anonymized post-mortems:
1. "The 80% pool-utilization warning alert (added after INC-2024-0089) fired at minute 2, giving on-call 6 minutes of headroom before the saturation alert at minute 8. The warning alert prevented a complete outage."
2. "The incident commander explicitly named a scribe at minute 4. The scribe's running log made the post-mortem timeline trivial to reconstruct and reduced post-mortem authoring time by an estimated 4 hours."
3. "Three engineers from different teams joined within 2 minutes of the page. The cross-team on-call rotation, despite being controversial when introduced, paid off — having a database engineer on the call from minute 3 enabled mitigation at minute 14 rather than minute 30+."
4. "Customer support's standard incident comm template (introduced in Q1) was published to the status page within 6 minutes of detection. Enterprise customers reported the proactive comm before they had noticed the outage."
5. "The runbook entry 'How to drain a node safely' was used directly, copy-pasted, and worked first try. The runbook had been reviewed 3 weeks prior."
6. "An on-call engineer pushed back on the suggested 'just restart Postgres' option at minute 11 and asked for 2 minutes to verify the hypothesis. The verification revealed the connection-leak source, which made the targeted mitigation possible. A blind restart would have masked the cause and likely caused a recurrence."
7. "The feature-flag rollback (`paymentApi.v2`) worked as designed at minute 17, ending customer impact within 30 seconds of the flag flip."
## When the section is truly empty
If after honest reflection the team cannot find 3 things that went well, that is a finding. Possibilities:
- The incident was the team's first of its kind. (Then "what went well" might be "the team learned a lot, and the documentation produced will help next time".)
- The team is exhausted and underinvested in tooling. (Then the post-mortem should surface this as a contributing factor and action item.)
- Psychological safety is too low for honest reflection. (Then escalate the cultural concern separately.)
Do not publish a post-mortem with an empty "What went well" section without explicitly noting why it is empty and what the team is doing about it.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| receiving-code-reviewobra/superpowersUse when receiving code review feedback, before implementing suggestions, especially if feedback seems unclear or technically questionable - requires technical rigor and verification, not performative agreement or blind implementation | Stars: 323k | Updated | Claude CodeCodex | Verified | Install |
| requesting-code-reviewobra/superpowersUse when completing tasks, implementing major features, or before merging to verify work meets requirements | Stars: 323k | Updated | Claude CodeCodex | Verified | Install |
| using-git-worktreesobra/superpowersUse when starting feature work that needs isolation from current workspace or before executing implementation plans - ensures an isolated workspace exists via native tools or git worktree fallback | Stars: 323k | Updated | Claude CodeCodex | Verified | Install |
| verification-before-completionobra/superpowersUse when about to claim work is complete, fixed, or passing, before committing or creating PRs - requires running verification commands and confirming output before making any success claims; evidence before assertions always | Stars: 323k | Updated | Claude CodeCodex | Verified | Install |