From ai-tools
Systematically investigates bugs, failures, and unexpected behavior across code, data, dependencies, and logs. Produces a structured analysis surfacing assumptions, evidence, gaps, and the simplest path forward.
How this skill is triggered — by the user, by Claude, or both
Slash command
/ai-tools:root-causeThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Structured investigation workflow for understanding problems completely before proposing fixes. This skill produces a detailed analysis document that:
Structured investigation workflow for understanding problems completely before proposing fixes. This skill produces a detailed analysis document that:
Announce at start: "I'm using the root-cause-analysis skill to investigate this problem systematically."
Critical Rule: Do NOT write, edit, or suggest any code changes. This is a thinking-only skill that surfaces what to fix, not how to fix it.
✅ Investigating unexpected behavior or errors in production or development
✅ Understanding why a test is failing repeatedly
✅ Diagnosing performance degradation or system failures
✅ Analyzing data corruption or inconsistent state
✅ Understanding third-party library issues or integration failures
✅ Tracing complex issues that span multiple services or layers
❌ Code review or style improvements (use relevant code standards)
❌ Feature development without a specific problem (use feature-writer)
Ask the user to describe:
Synthesize into a one-sentence problem statement: name the behavior, context, and expected vs actual outcome.
State this back to the user for confirmation.
Review the entire conversation history for:
Create a summary of what's been reported and flag anything unclear or contradictory.
Read every relevant file, not just the ones mentioned:
git log --oneline -20 and git diff for affected files)Document file paths and line numbers specifically — vague analysis is useless analysis.
Examine actual runtime values, not just the code:
For each data point, document:
Identify every external and internal dependency involved:
Document each as a table:
| Dependency | Type | Version | Role in Problem | Status |
|---|---|---|---|---|
| name | package / crate / tool / service | version | what it does | OK / Suspect / Unverified |
For files involved in the problem:
git log --oneline -20 -- <filename> for recent changesgit show <sha> to examine what changedRequest relevant logs from the user and examine for error messages, timeouts, and unexpected state transitions.
Ask:
State the problem in one sentence — name the behavior, context, and expected vs actual outcome.
List every assumption with confidence level and basis.
List every approach or fix that has been tried, with results.
Describe actual data at each stage of the problem.
Table of all external and internal dependencies.
List concrete evidence with interpretations and certainty levels.
List what you don't know but need to know, how to get it, and why it matters.
Rank candidate root causes by probability with key evidence.
For each candidate root cause, describe the simplest possible fix.
After completing the analysis:
The user decides what to do with your analysis.
Provides CDSS development patterns for drug interaction checking, dose validation, clinical scoring (NEWS2, qSOFA), and alert classification integrated into EMR workflows.
npx claudepluginhub sean-m-cooper/ai_tools --plugin ai-tools