postmortem
Skill1 starsChanged 6 days agoMITView on GitHub
Write a blameless incident postmortem collaboratively. Draft from a completed investigation, then iterate with the engineer section by section. It will NOT let someone offload the whole thing for anything beyond a trivial incident, because a real postmortem needs their timeline, detection story, severity and owners. Use for "write a postmortem", "write up the incident", "do an RCA writeup", or after an investigation concludes.
45 other projects use a version of this.
Adopted by
Most recently taken up by
- ychampion/ystacktheir own version, first seen Nov 15, 2026
- Dimon94/dveritytheir own version, first seen Nov 15, 2026
- Dimon94/skillstheir own version, first seen Nov 15, 2026
- mehboobali98/rmine-skillstheir own version, first seen Nov 6, 2026
- SethGammon/Citadeltheir own version, first seen Nov 4, 2026
- flonat/flonat-researchtheir own version, first seen Nov 3, 2026
Used with
Projects with it also use hypothesis-driven-debugging, status-report and 10 more.
Other files people use alongside this one.
Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
Sturla22Skillhypothesis-driven-debugging
Debug by ranking concrete hypotheses and running discriminating checks. Use when a bug is hard to localize, many plausible failure points exist, logs are noisy or misleading, a symptom appears intermittently, or trial-and-error edits haven't worked.
Also in 3 projects that use this
AnthropicSkillstatus-report
Generate a status report with KPIs, risks, and action items. Use when writing a weekly or monthly update for leadership, summarizing project health with green/yellow/red status, surfacing risks and decisions that need stakeholder attention, or turning a pile of project tracker activity into a readable narrative.
Also in 3 projects that use this
seongjaeryuSkillorganize
Organize knowledge into Sets
Also in 3 projects that use this
SaluteYBSkillskill-eval
Evaluate, tune, and regression-check AgentSkills. Use when creating a new skill or improving an existing one and you need structured should-trigger, should-not-trigger, boundary-case, and execution-quality examples. Also use when auditing overlapping skills and sharpening trigger descriptions before release.
Also in 3 projects that use this
neurofooSkillpremortem
Pre-mortem analysis that imagines a plan has failed, then works backward to identify causes and preventions. Use before launches, major decisions, or risky initiatives to surface hidden risks.
Also in 3 projects that use this
AnthropicSkillincident-response
Run an incident response workflow — triage, communicate, and write postmortem. Trigger with "we have an incident", "production is down", an alert that needs severity assessment, a status update mid-incident, or when writing a blameless postmortem after resolution.
Also in 4 projects that use this
Other versions45 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 45
- dotnet/fsharp★ 4.3ksame name
- SethGammon/Citadel★ 927same name
- boshu2/agentops★ 454same name
- flonat/flonat-research★ 150same name
- neurofoo/agent-skills★ 111same name
- Dimon94/dverity★ 107same name
- Dimon94/skills★ 107same name
- EthanReedLabs/sulde★ 99same name
- Hongyu-yu/matsci-ai-skills★ 19same name
- Hassaan146/claude-skills★ 18same name
35 more not listed
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository33 other files it gives its agents.
- AGENTS.mdAGENTS.md
- hands-offSkill
- hill-climbSkill
- optimize-agent-instructionsSkill
- asset-gardenSkill
- clean-code-stewardSkill
- css-override-censusSkill
- debt-ledgerSkill
- find-component-reinventionSkill
- find-contract-driftSkill
- find-dead-codeSkill
- find-dead-cssSkill
- find-dependency-overlapSkill
- find-reinventionSkill
- find-token-driftSkill
- optimize-commentsSkill
- optimize-docsSkill
- pattern-censusSkill
- test-gardenSkill
- theme-paritySkill
- type-rhythmSkill
- vocabulary-driftSkill
- ui-feature-specSkill
- oie-channel-code-reviewSkill
- oie-plugin-developmentSkill
- batch-workSkill
- deep-code-reviewSkill
- triage-cardSkill
- write-testsSkill
- brainstormSkill
- debate-planSkill
- steelmanSkill
- unslopSkill
The file
We split this skill into the rules it gives an agent
8 rules. 0 are common, 8 are distinctive, and 0 are very specific to this project.
0
Common
Most of the 41 edited versions we read keep these.
8
Distinctive
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 31
If they say "just write the postmortem", push back for anything beyond a trivial incident.
- Distinctive: line 35
Only a genuinely simple, low-impact incident may be drafted end to end with one review pass.
- Distinctive: line 67
Write in the template's order.
- Distinctive: line 72
Detection: Write what they told you, not "a human noticed".
- Distinctive: line 88
Only call it done when they have reviewed and accepted it, with any requested corrections incorporated.
- Distinctive: line 90
Leave the commit and push to them.
- Distinctive: line 98
Evidence over narrative: Every claim of cause traces to a log, metric or code citation.
- Distinctive: line 100
Do not finalize unilaterally: If they have not reviewed it, it is a draft, and it says so at the top.
“Common” is measured: kept by at least 6 in 10 of the 41 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Postmortem, collaborative and blameless
Turn a completed, verified investigation into a postmortem that follows best practice. The section template, folder convention and card format are in reference/TEMPLATE.md.
A postmortem is a conversation, not a deliverable you hand back. The engineer was there and you were not. Your draft is a scaffold built from investigation evidence. They fill the gaps, correct what you inferred, and own the narrative. Do not write the whole thing and declare it done.
Two hard rules
1. Back-and-forth is required
Draft a section, or a skeleton, then stop and ask them to review and correct it before moving on. Explicitly surface what you do not know and what you assumed.
Expect several rounds. That back-and-forth is the point, not friction.
2. Do not enable offloading
If they say "just write the postmortem", push back for anything beyond a trivial incident. Ask for the things only they know rather than inventing them. A confidently wrong postmortem is worse than a thin one, because it will be read by people who were not there and cannot tell.
Only a genuinely simple, low-impact incident may be drafted end to end with one review pass.
What you must get from them, and must not invent
- Detection. How it was actually noticed. A support ticket? An alert? A customer? Stumbled on during unrelated work? This is almost always different from "we saw the metric", and it is where the process lessons live. Ask.
- Timeline. The real sequence and times, especially the firefight, which is often a different day from the analysis. Confirm the order of events.
- Severity and impact. Business impact, who and what was affected, duration, any data concerns. They set severity, not you.
- Owners and feasibility. Who owns each follow-up, and which fixes are actually doable versus blocked on a vendor or another team.
- Corrections. Anything in your draft they know is wrong. Invite this every round.
Workflow
1. Ground it in the investigation
Base the draft on a verified root cause with an evidence chain. If the cause is not actually nailed down, say so and do not write a postmortem yet. Investigate first.
A prior investigation's disproven hypotheses map onto a section of the template, and they stop the next person re-walking the same wrong turns.
2. Draft the skeleton, confirm scope
Propose the incident title, dates, severity, affected systems and a one-line summary. Get those confirmed or corrected before fleshing anything out. Create the file early so edits land in place.
3. Fill sections, pausing on the high-stakes ones
Write in the template's order. After each of these, stop and ask:
- Executive summary. Plain language. Someone senior can read only this and stop. Confirm the framing.
- Timeline. The section most likely to be wrong without them. Draft from evidence, then ask what is out of order, missing or mis-timed.
- Detection. Write what they told you, not "a human noticed".
- Root cause. Symptom versus cause, contributing factors, five whys, and a disproven-hypotheses subsection. Cite code and logs.
- Action items. Bucketed by how much control the team actually has, each owned and prioritised. Mark what is already done.
Flag inline what you inferred versus what is evidenced. Leave explicit markers for gaps you need them to fill, so an unfinished draft cannot be mistaken for a finished one.
4. Offer cards for the action items
Offer to draft issue-tracker cards, one per action-item bucket as they direct: a feature card for the code fixes, spike cards for the uncertain ones. Put them beside the postmortem.
5. Finalize
Only call it done when they have reviewed and accepted it, with any requested corrections incorporated. Confirm: no customer data, no secrets, severity agreed, owners assigned, action items actionable.
Leave the commit and push to them.
Guardrails
- Blameless. Failures are of design, guardrails and observability, never of people. No names as causes.
- No customer data, no secrets. In a regulated domain, reference records by internal identifier rather than anything that identifies a person. Never paste credentials. Scrub the examples too.
- Evidence over narrative. Every claim of cause traces to a log, metric or code citation. "We think" is fine when labelled. Do not dress inference as fact.
- Do not finalize unilaterally. If they have not reviewed it, it is a draft, and it says so at the top.
It is working if
Confirm the engineer reviewed and explicitly accepted the timeline, detection story, severity, and action owners, with any corrections incorporated. Check each causal claim against its cited evidence, and retain disproven hypotheses when the investigation had them. Approval without corrections is valid. Unanswered questions remain marked as gaps, and the document stays a draft until the engineer accepts it.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

