Claude Code Plugins

Community-maintained marketplace

Feedback

Apply Edward de Bono's Six Thinking Hats methodology to software testing for comprehensive quality analysis. Use when designing test strategies, conducting test retrospectives, analyzing test failures, evaluating testing approaches, or facilitating testing discussions. Each hat provides a distinct testing perspective: facts (White), risks (Black), benefits (Yellow), creativity (Green), emotions (Red), and process (Blue).

Install Skill

1Download skill
2Enable skills in Claude

Open claude.ai/settings/capabilities and find the "Skills" section

3Upload to Claude

Click "Upload skill" and select the downloaded ZIP file

Note: Please verify skill by going through its instructions before using it.

SKILL.md

name six-thinking-hats
description Apply Edward de Bono's Six Thinking Hats methodology to software testing for comprehensive quality analysis. Use when designing test strategies, conducting test retrospectives, analyzing test failures, evaluating testing approaches, or facilitating testing discussions. Each hat provides a distinct testing perspective: facts (White), risks (Black), benefits (Yellow), creativity (Green), emotions (Red), and process (Blue).
category methodology
priority medium
tokenEstimate 1100
agents qe-quality-analyzer, qe-regression-risk-analyzer, qe-test-generator
implementation_status optimized
optimization_version 1
last_optimized Wed Dec 03 2025 00:00:00 GMT+0000 (Coordinated Universal Time)
dependencies
quick_reference_card true
tags thinking, methodology, decision-making, collaboration, analysis

Six Thinking Hats for Testing

When analyzing testing decisions: 1. DEFINE focus clearly (specific testing question) 2. APPLY each hat sequentially (5 min each) 3. DOCUMENT insights per hat 4. SYNTHESIZE into action plan

Quick Hat Rotation (30 min):

🤍 WHITE (5 min) - Facts only: metrics, data, coverage
❤️ RED (3 min) - Gut feelings (no justification needed)
🖤 BLACK (7 min) - Risks, gaps, what could go wrong
💛 YELLOW (5 min) - Strengths, opportunities, what works
💚 GREEN (7 min) - Creative ideas, alternatives
🔵 BLUE (3 min) - Action plan, next steps

Example for "API Test Strategy":

  • 🤍 47 endpoints, 30% coverage, 12 integration tests
  • ❤️ Anxious about security, confident on happy paths
  • 🖤 No auth tests, rate limiting untested, edge cases missing
  • 💛 Good docs, CI/CD integrated, team experienced
  • 💚 Contract testing with Pact, chaos testing, property-based
  • 🔵 Security tests first, contract testing next sprint

Quick Reference Card

The Six Hats

Hat Focus Key Question
🤍 White Facts & Data What do we KNOW?
❤️ Red Emotions What do we FEEL?
🖤 Black Risks What could go WRONG?
💛 Yellow Benefits What's GOOD?
💚 Green Creativity What ELSE could we try?
🔵 Blue Process What should we DO?

When to Use Each Hat

Hat Use For
🤍 White Baseline metrics, test data inventory
❤️ Red Team confidence check, quality gut feel
🖤 Black Risk assessment, gap analysis, pre-mortems
💛 Yellow Strengths audit, quick win identification
💚 Green Test innovation, new approaches, brainstorming
🔵 Blue Strategy planning, retrospectives, decision-making

Hat Details

🤍 White Hat - Facts & Data

Output: Quantitative testing baseline

Questions:

  • What test coverage do we have?
  • What is our pass/fail rate?
  • What environments exist?
  • What is our defect history?
Example Output:
Coverage: 67% line, 45% branch
Test Suite: 1,247 unit, 156 integration, 23 E2E
Execution Time: Unit 3min, Integration 12min, E2E 45min
Defects: 23 open (5 critical, 8 major, 10 minor)

🖤 Black Hat - Risks & Cautions

Output: Comprehensive risk assessment

Questions:

  • What could go wrong in production?
  • What are we NOT testing?
  • What assumptions might be wrong?
  • Where are the coverage gaps?
HIGH RISKS:
- No load testing (production outage risk)
- Auth edge cases untested (security vulnerability)
- Database failover never tested (data loss risk)

💛 Yellow Hat - Benefits & Optimism

Output: Strengths and opportunities

Questions:

  • What's working well?
  • What strengths can we leverage?
  • What quick wins are available?
STRENGTHS:
- Strong CI/CD pipeline
- Team expertise in automation
- Stakeholders value quality

QUICK WINS:
- Add smoke tests (reduce incidents)
- Automate manual regression (save 2 days/release)

💚 Green Hat - Creativity

Output: Innovative testing ideas

Questions:

  • How else could we test this?
  • What if we tried something completely different?
  • What emerging techniques could we adopt?
IDEAS:
1. AI-powered test generation
2. Chaos engineering for resilience
3. Property-based testing for edge cases
4. Production traffic replay
5. Synthetic monitoring

❤️ Red Hat - Emotions

Output: Team gut feelings (NO justification needed)

Questions:

  • How confident do you feel about quality?
  • What makes you anxious?
  • What gives you confidence?
FEELINGS:
- Confident: Unit tests, API tests
- Anxious: Authentication flow, payment processing
- Frustrated: Flaky tests, slow E2E suite

🔵 Blue Hat - Process

Output: Action plan with owners and timelines

Questions:

  • What's our strategy?
  • How should we prioritize?
  • What's the next step?
PRIORITIZED ACTIONS:
1. [Critical] Address security testing gap - Owner: Alice
2. [High] Implement contract testing - Owner: Bob
3. [Medium] Reduce flaky tests - Owner: Carol

Session Templates

Solo Session (30 min)

# Six Hats Analysis: [Topic]

## 🤍 White Hat (5 min)
Facts: [list metrics, data]

## ❤️ Red Hat (3 min)
Feelings: [gut reactions, no justification]

## 🖤 Black Hat (7 min)
Risks: [what could go wrong]

## 💛 Yellow Hat (5 min)
Strengths: [what works, opportunities]

## 💚 Green Hat (7 min)
Ideas: [creative alternatives]

## 🔵 Blue Hat (3 min)
Actions: [prioritized next steps]

Team Session (60 min)

  • Each hat: 10 minutes
  • Rotate through hats as group
  • Document on shared whiteboard
  • Blue Hat synthesizes at end

Agent Integration

// Risk-focused analysis (Black Hat)
const risks = await Task("Identify Risks", {
  scope: 'payment-module',
  perspective: 'black-hat',
  includeMitigation: true
}, "qe-regression-risk-analyzer");

// Creative test approaches (Green Hat)
const ideas = await Task("Generate Test Ideas", {
  feature: 'new-auth-system',
  perspective: 'green-hat',
  includeEmergingTechniques: true
}, "qe-test-generator");

// Comprehensive analysis (All Hats)
const analysis = await Task("Six Hats Analysis", {
  topic: 'Q1 Test Strategy',
  hats: ['white', 'black', 'yellow', 'green', 'red', 'blue']
}, "qe-quality-analyzer");

Agent Coordination Hints

Memory Namespace

aqe/six-hats/
├── analyses/*        - Complete hat analyses
├── risks/*           - Black hat findings
├── opportunities/*   - Yellow hat findings
└── innovations/*     - Green hat ideas

Fleet Coordination

const analysisFleet = await FleetManager.coordinate({
  strategy: 'six-hats-analysis',
  agents: [
    'qe-quality-analyzer',        // White + Blue hats
    'qe-regression-risk-analyzer', // Black hat
    'qe-test-generator'           // Green hat
  ],
  topology: 'parallel'
});

Related Skills


Anti-Patterns

❌ Avoid Why ✅ Instead
Mixing hats Confuses thinking One hat at a time
Justifying Red Hat Kills intuition State feelings only
Skipping hats Misses insights Use all six
Rushing Shallow analysis 5 min minimum per hat

Remember

Separate thinking modes for clarity. Each hat reveals different insights. Red Hat intuition often catches what Black Hat analysis misses.

Everyone wears all hats. This is parallel thinking, not role-based. The goal is comprehensive analysis, not debate.