root-cause-analysis（根本原因分析）

Root Cause Analysis Orchestration Skill

This skill helps you systematically identify the root cause of any problem using proven methodologies from the Toyota Production System and other industry-standard techniques.

Quick Reference: When to Load Which Resource

Core Principle

Do not treat symptoms—find and fix the root cause. As Taiichi Ohno, architect of the Toyota Production System, said: "By repeating why five times, the nature of the problem as well as its solution becomes clear."

Orchestration Protocol

Phase 1: Problem Classification

Quickly identify your problem domain and complexity:

Problem Domain:

• Software: Code bugs, system failures, performance, deployment
• Hardware: Equipment, mechanical, electrical, maintenance
• Process: Workflow, procedures, organizational, communication
• Personal: Life challenges, productivity, habits, wellbeing

Complexity Level:

• Simple: Clear failure chain, 1-2 likely causes → Use 5 Whys
• Complex: Multiple possible causes, unknown scope → Start with Fishbone
• Critical/Safety: High stakes, needs rigor → Use Fault Tree
• Multiple Issues: Many competing problems → Use Pareto first

Action: Load appropriate resource file(s) based on classification.

Phase 2: Methodology Selection

Based on problem type, select your approach:

Phase 3: Execution & Verification

During Analysis:

1. Define problem clearly (What/Where/When/Impact)
2. Gather evidence systematically
3. Apply selected methodology
4. Document reasoning at each step
5. Verify root cause with Forward/Backward tests

Before Finalizing:

• Validate conclusion against evidence
• Check for red flags (see common-root-causes.md)
• Confirm actionability (can you fix this?)
• Develop solutions addressing root cause

Problem Definition Framework

Create a clear problem statement before analysis:

Essential Elements:

• What: Observable symptom (not assumed cause)
• Where: Location/system/component affected
• When: Timeline, frequency, pattern
• Impact: Users/systems affected, severity

Example: "Users in EU region experience 3-5 second dashboard load delays during 9-11 AM UTC peak hours, affecting ~2,000 daily active users. Started after v2.4 deployment on Nov 18th."

Evidence Gathering (Go and See)

Follow Toyota's principle—collect facts, not opinions:

Key Evidence Sources:

• Logs, metrics, monitoring data
• Timeline of events and changes
• System/code/configuration changes before problem
• Environmental factors (load, traffic, season)
• User reports and reproduction steps
• System state before/during/after

RCA Methodologies

See resources/rca-methodologies.md for complete methodology guide.

Resource Files Summary

resources/rca-methodologies.md

Comprehensive methodology guide covering:

• 5 Whys: Step-by-step process with software examples
• Fishbone Diagram: Structure, 6 M's categories, process
• Pareto Analysis: Prioritization using 80/20 rule
• Fault Tree Analysis: Top-down formal analysis
• Barrier Analysis: Control failure examination
• Structured 6-phase RCA process, domain-specific guidance, templates

resources/common-root-causes.md

Pattern reference catalog by domain:

• Software Engineering: Code defects, configuration, dependencies, deployment
• Hardware & Equipment: Mechanical, electrical, operational, maintenance
• Process & Operations: Workflow, design, resources
• Personal/Life: Health, habits, environment, skills
• Red flags, recurring themes, pattern recognition

resources/example-analyses.md

Four worked examples with full analysis:

1. Software Bug: JWT authentication (5 Whys)
2. Vehicle Maintenance: Overheating (5 Whys)
3. System Failure: E-commerce checkout (Fishbone + 5 Whys)
4. Personal Productivity: Missed deadlines (Fishbone + 5 Whys)

resources/advanced-techniques.md

Formal methods for complex problems:

• Fault Tree Analysis: Boolean logic, safety systems
• Barrier Analysis: Control failures
• Multi-Methodology Chains: Complex orchestration
• Verification Frameworks: Comprehensive testing

How This Skill Works

1. Clarify your situation: Domain, observations, evidence, time
2. Recommend approach: Complexity analysis, methodology, resources
3. Guide through analysis: Problem statement, evidence, methodology, verification
4. Deliver output: Analysis, root cause, solutions, implementation

Quick Start: 5-Minute RCA

1. State problem (What/Where/When/Impact)
2. First Why: fact-based answer
3. Second Why: dig deeper
4. Third Why: dig deeper again
5. Verify: would fixing this prevent it?

Templates & Examples

• 5 Whys Template in resources/rca-methodologies.md
• Fishbone Template in resources/rca-methodologies.md
• Worked Examples in resources/example-analyses.md
• Solution Structures in resources/example-analyses.md

Next Steps

1. Identify problem domain (software/hardware/process/personal)
2. Load appropriate resource from table above
3. Select methodology based on complexity
4. Follow step-by-step process in resource
5. Verify root cause (Forward/Backward tests)
6. Develop actionable solutions

Remember: Goal is systematic investigation—disciplined questioning until you reach a cause you can actually fix.