AgentSkillsCN

rca-investigation

针对Bug与突发事件开展根本原因分析。在调查故障、回归问题，或应对意外行为时使用。

SKILL.md
--- frontmatter
name: rca-investigation
description: Conduct root cause analysis for bugs and incidents. Use when investigating failures, regressions, or unexpected behavior.
metadata:
  version: '1.0'

Root Cause Analysis

Investigation Process

1. Reproduce

  • •Confirm the issue is reproducible
  • •Document exact steps, inputs, and environment
  • •Note whether the issue is deterministic or intermittent

2. Isolate

  • •Identify the minimal reproduction case
  • •Narrow to the specific module, function, or data condition
  • •Use binary search (git bisect, code commenting) to find the introducing change

3. Root Cause

  • •Identify why the failure occurs, not just what fails
  • •Distinguish root cause from symptoms and contributing factors
  • •Check if the root cause affects other code paths

4. Recommend

  • •Propose a fix approach
  • •Identify which agent should implement the fix
  • •Estimate impact and risk of the fix

5. Prevent

  • •Recommend specific tests to prevent recurrence
  • •Suggest architectural changes if the root cause is systemic
  • •Recommend monitoring or assertions for early detection

Report Template

markdown
# RCA: [Brief descriptive title]

**Date**: YYYY-MM-DD
**Severity**: Critical | High | Medium | Low
**Status**: Investigating | Root Cause Identified | Fix Verified

## Issue Summary

One-paragraph description of what was observed.

## Reproduction

1. Step-by-step reproduction instructions
2. ...

## Timeline

- [timestamp] Issue first observed
- [timestamp] Investigation began
- [timestamp] Root cause identified

## Root Cause

Detailed explanation of why the issue occurred.

## Contributing Factors

- Factor 1
- Factor 2

## Fix Recommendation

What should be changed, by which agent, with what priority.

## Prevention Measures

- Tests to add
- Checks to implement
- Architectural changes to consider