Guide · Agent skill
How to use the Root Cause Analysis agent skill
Updated · written with AI from the project's own documentation and checked against our catalogue data
RootCauseAnalysis is an agent skill: a folder of instructions that a SKILL.md-compatible agent such as Claude Code reads when you ask it to investigate a failure. It is aimed at engineers writing postmortems and teams dealing with bugs or failures that keep returning. After this guide you will know what the skill changes about the agent's behaviour, how to install it, how to phrase requests so it triggers, and what evidence to supply so the output is useful.
Root Cause Analysisdanielmiessler/LifeOS · MIT Download the skill Quick start
- 1Download the ZIP. Download the RootCauseAnalysis skill ZIP from this page.
- 2Unzip the skill. Unzip it so the folder contains SKILL.md and the Intents folder.
- 3Place it where your agent looks. Put it in ~/.claude/skills/ for personal use or in the project's .claude/skills/ folder, or upload it under Settings → Capabilities → Skills in the Claude apps.
- 4Gather your evidence. Collect logs, timestamps, the timeline of changes and the observed symptoms of the failure.
- 5Ask with a natural request. Describe the failure and use a trigger phrase such as root cause, postmortem or 5 whys.
- 6Review the gaps and actions. Check which causes are marked as hypotheses or gaps and whether each corrective action addresses a systemic cause.
What the RootCauseAnalysis skill makes the agent do
Without the skill, an agent asked why something failed tends to offer a quick guess. With it installed, the agent treats the failure as an investigation. It chooses a structured method that fits the problem, such as asking 'why' repeatedly, sorting causes into categories, building a fault tree, or writing a blameless postmortem, and it works from the evidence you provide.
The skill is designed to name systemic causes rather than blame a person, and to flag gaps in the evidence instead of inventing explanations. It covers five intents: Five Whys, Fishbone, Postmortem, Fault Tree and Kepner-Tregoe.
- Five Whys: a quick causal chain for a single failure point.
- Fishbone: a categorised cause map when several categories are suspected, such as people, process and tools.
- Postmortem: a formal, blameless review for production outages or security incidents.
- Fault Tree: top-down Boolean analysis for complex or safety-critical failures.
- Kepner-Tregoe: distinction analysis for subtle defects, phrased as 'why here and not there?'.
Install the skill from the ZIP
Download the ZIP of the RootCauseAnalysis skill from this page and unzip it. The skill folder must contain the SKILL.md file and its Intents, MethodSelection.md and Foundation.md files.
For a personal install, unzip it into ~/.claude/skills/. For a single project, unzip it into the project's .claude/skills/ folder. If you use the Claude apps instead, upload the skill under Settings → Capabilities → Skills.
- Personal install: ~/.claude/skills/
- Project install: the project's .claude/skills/ folder
- Claude apps: Settings → Capabilities → Skills
Trigger the skill with natural requests
You do not need a special command. The skill's description lists the phrases that should route to it, including root cause, RCA, 5 whys, fishbone, postmortem, incident analysis, fault tree, why does this keep failing, blameless and recurring bug. Use them naturally and include the evidence you have.
The skill's own examples show the pattern: name the failure, say what you want, and give the context.
- 'Payments threw 500s for ten minutes after last night's deploy. Write the postmortem.'
- 'The export job fails only on the EU cluster and only on Mondays. Why?'
- '5 whys on why the nightly backup keeps missing its window'
Supply the evidence the analysis needs
The skill works from the evidence you supply: logs, timelines and descriptions of the failure. Its guidance says conclusions must rest on facts you have supplied, and that a hypothesis is not treated as proven. Paste the relevant log lines, timestamps and what changed before the failure, and say which parts you are unsure of.
If evidence is missing, the skill is meant to mark those points as gaps or hypotheses. Expect that output rather than a confident single cause when the data is thin.
- Logs or error messages with timestamps
- A timeline of deploys, config changes and alerts
- The observed symptoms and when they started and stopped
- Anything you already suspect, labelled as a suspicion
Know the limits and safety notes
The skill's files do not run scripts and do not use the network. The source does not specify any API keys or accounts. It requires Claude Code or another SKILL.md-compatible agent, plus the incident evidence you provide.
The skill is not for feedback loops or long-term structural patterns; its own description says to use the separate SystemsThinking skill for those. It is also not suited to live incidents that need an answer right now, because a full analysis takes time and evidence gathering.
- Runs scripts: no
- Uses network: no
- Not for systemic loops: use SystemsThinking
- Not for urgent live incidents needing an immediate answer
Tips for better results
Give the agent the full timeline, not only the final error. The skill's Postmortem intent expects a timeline and a map of defensive layers, so supplying those up front gives it more to work with.
Check the corrective actions it proposes. The skill aims for actions that prevent a class of recurrence rather than fixing only the latest trigger, and it ends at causes that are within the team's authority to change.
- Start with a single failure for Five Whys and a broader incident for Postmortem.
- Ask for a fault tree when the failure has several paths that combine with AND/OR logic.
- Ask the agent to list its gaps if you want to see what the evidence does not support.
What people use it for
- Writing a blameless postmortem after a production outage
- Finding why a bug keeps returning after each fix
- Running a quick Five Whys on a single failing job or process
- Mapping several suspected cause categories with a fishbone diagram
- Working through a distinction question such as why a failure happens on one cluster and not another
Frequently asked questions
Is the RootCauseAnalysis skill free?
The source files are published under the MIT licence, which is stated on the item page. The source does not mention any price, so check the licence text for the terms that apply to you.
What do I need to use the RootCauseAnalysis skill?
You need Claude Code or another SKILL.md-compatible agent. You also need incident evidence such as logs, timelines and descriptions of the failure. The source does not specify any API keys or accounts.
Where do I install the skill in Claude Code?
Unzip the skill into ~/.claude/skills/ for a personal install, or into a project's .claude/skills/ folder for one project. In the Claude apps, upload it under Settings → Capabilities → Skills.
Does the skill run scripts or access the internet?
According to the item details, it runs no scripts and uses no network access.
Which skill should I use for feedback loops?
Use the separate SystemsThinking skill for feedback loops and long-term structural patterns. The RootCauseAnalysis skill says it is not built for those.
Does the skill blame people for failures?
No. The skill is designed to trace failures to systemic root causes and to write postmortems in a blameless way, though it still depends on the quality of the evidence you supply.
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
RootCauseAnalysis suits engineers writing postmortems after outages or security incidents, and teams trying to stop a failure that keeps coming back. It is a poor fit for urgent live incidents that need an immediate answer, for feedback-loop analysis, or for questions that depend on facts you cannot supply or check. If you want a fast guess, a plain prompt will do; if you want an evidence-based investigation, this skill is built for that.
Download the skill