Guide · Agent skill
How to use the Post-Mortem Analyst agent skill (postmortem)
Updated · written with AI from the project's own documentation and checked against our catalogue data
The postmortem skill is a text-based analysis framework for AI coding and chat agents that support skills. When you ask the agent to review a failure, it works through a structured, blame-free method: what was expected versus what happened, a 5 Whys chain, root causes separated from contributing factors, ignored warning signs, and a change register with owners and dates. This guide explains what the skill does, how to install it, how to trigger it, and where it falls short.
Post-Mortem Analystalirezarezvani/claude-skills · MIT Download the skill Quick start
- 1Download the ZIP. Download the postmortem skill ZIP from this page. It contains SKILL.md.
- 2Unzip into a skills folder. Unzip it into ~/.claude/skills/ for personal use, or into .claude/skills/ inside a project.
- 3Or upload it in the Claude apps. Go to Settings → Capabilities → Skills and upload the ZIP.
- 4Describe the event. Write the expected outcome, the actual outcome, the dates and the impact in as much detail as you have.
- 5Run the command. Use /em:postmortem <event> or ask in plain language for a blameless 5 Whys analysis.
- 6Review the root cause and change register. Check that the root cause is distinct from contributing factors and that each action has an owner, a due date and a verification method.
- 7Set the verification date. Schedule the check-in date the skill proposes and confirm whether the changes happened and worked.
What the postmortem skill makes the agent do
The skill is built around the command /em:postmortem <event>. It is meant for failures such as a failed launch, a missed quarter, a deal that fell through, or a hire that didn't work out. The goal is to understand what actually happened and why, not to decide whose fault it was.
The skill guides the agent through seven steps: define the event precisely, run a proper 5 Whys chain, separate contributing factors from the root cause, identify warning signs that were ignored, sort causes into what was in and out of control, build a change register, and set a verification date. The agent then fills in a fixed output format with event, expected and actual outcomes, timeline, 5 Whys, root cause, contributing factors, missed warning signs, a change register table, and a verification date.
- It is a thinking framework. The skill does not look up any data itself, so you supply the facts about the event.
- It contains no technical diagnostic tools, so it is not suited to live debugging such as production outages.
Install the postmortem skill from the ZIP
Download the ZIP for the skill from this page. The package contains a single file, SKILL.md. You can install it in one of two ways.
For a personal install, unzip the file into ~/.claude/skills/. For a single project, unzip it into the .claude/skills/ folder inside that project. The skill folder should contain SKILL.md.
You can also upload the ZIP in the Claude apps under Settings → Capabilities → Skills.
- Personal: ~/.claude/skills/
- Project: .claude/skills/ inside your project folder
- Claude apps: Settings → Capabilities → Skills, then upload the ZIP
Trigger the skill with a natural request
You can invoke the skill with the command /em:postmortem followed by a short description of the event. You can also ask in plain language after a failure, such as asking the agent to analyze why a launch went wrong. Include the facts you have: the expected result, the actual result, dates, and the impact.
The skill works best when you give it concrete numbers and a timeline. Vague input produces vague analysis, because the first step of the framework is to pin down the expected and actual outcomes.
/em:postmortem <event>Examples of good requests
A useful request names the event, the gap, and what you know about it. The source example is dissecting why a Q3 release slipped six weeks. A revenue example from the framework is a Q3 miss where deals slipped out of quarter and one was lost to a competitor.
For a bad hire, give the role, the timeline of the hiring process, what the interviewers saw, and any concerns that were raised along the way. The framework will then look for the systemic cause, such as an undefined competency framework.
- /em:postmortem why the Q3 release slipped six weeks
- /em:postmortem the enterprise launch that missed its pipeline target, with the deal list and dates
- /em:postmortem the senior hire who left after four months, with the interview notes and the concerns raised
What the skill needs and safety notes
The skill lists no requirements. It does not run scripts and does not use the network. Because it is a text framework, everything it works from is what you put into the conversation, so avoid pasting confidential personnel details into tools that you are not allowed to use for that purpose.
The skill is an analysis aid, not an authoritative inquiry. It is not suitable for legal, HR disciplinary, or compliance investigations, which need specialist process and professional advice.
- Requires: nothing listed
- Runs scripts: no
- Uses network: no
Tips for a useful post-mortem
Write the expected and actual outcomes before you ask for analysis. The skill's first step depends on that precision. Then check the output: each change should name what changes, who owns it, the due date, and how you will verify it.
Use the verification date. The skill treats it as the most commonly skipped step, and without it the post-mortem is hard to act on.
- Replace vague actions like 'communication will be better' with a named owner and a date.
- Ask whether each cause is a root cause or a contributing factor. Fixing only contributing factors tends to repeat the same failure.
- Check the 'out of control' list. Saying something was outside your control can be a way to avoid accountability, so test it.
What people use it for
- Reviewing a failed product launch to find why it slipped and what process change prevents it.
- Analyzing a missed quarterly revenue target, separating deals lost to competitors from pipeline problems.
- Examining a bad hire to find systemic gaps such as an undefined competency framework.
- Running a team retrospective that produces owned, dated actions instead of vague commitments.
- Learning how a blameless root-cause analysis is structured by working through a real past event.
Frequently asked questions
Is the postmortem agent skill free?
The skill is released under the MIT licence, as listed on this page. Check the licence file in the source repository for the full terms.
What do I need to install the postmortem skill?
The skill lists no requirements. You download the ZIP, unzip it into ~/.claude/skills/ or a project's .claude/skills/ folder, or upload it in the Claude apps under Settings → Capabilities → Skills.
Does the postmortem skill run scripts or access the internet?
No. The skill runs no scripts and uses no network. It is a text-based framework that works only from the information you give it.
Can I use this for a production outage?
Not as a debugging tool. The skill contains no technical diagnostic tools, so it is meant for business failures such as launches, quarters and hires. Use a dedicated incident process for live technical problems.
Can I use it for an HR or legal investigation?
No. The skill is an analysis aid, not an authoritative inquiry. Legal, HR disciplinary and compliance investigations need specialist process and professional advice.
How is this different from a generic root-cause prompt?
It enforces a specific sequence: define the event precisely, build a real 5 Whys chain, separate root causes from contributing factors, check ignored warning signs, and end with owned, dated actions and a verification date.
New agent skills, repos and AI tools — every week
The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.
Our verdict
Use this skill if you are a leader or manager reviewing a failed launch, missed target or bad hire, or if you want retrospectives that produce concrete fixes. Pick something else for legal, HR disciplinary or compliance investigations, or for live technical incidents such as outages, since the skill is only a text-based analysis aid.
Download the skill