agentic-quality-engineering（Agent 质量工程）

Agentic Quality Engineering

<default_to_action> When implementing agentic QE or coordinating agents:

1. SPAWN appropriate agent(s) for the task using Task tool with agent type
2. CONFIGURE agent coordination (hierarchical/mesh/sequential)
3. EXECUTE with PACT principles: Proactive analysis, Autonomous operation, Collaborative feedback, Targeted risk focus
4. VALIDATE results through quality gates before deployment
5. LEARN from outcomes - store patterns in aqe/learning/* namespace

Quick Agent Selection:

• Test generation needed → qe-test-generator
• Coverage gaps → qe-coverage-analyzer
• Quality decision → qe-quality-gate
• Security scan → qe-security-scanner
• Performance test → qe-performance-tester
• Full pipeline → qe-fleet-commander

Critical Success Factors:

• Agents amplify human expertise, not replace it
• Human-in-the-loop for critical decisions
• Measure: bugs caught, time saved, coverage improved </default_to_action>

Quick Reference Card

When to Use

• Designing autonomous testing systems
• Scaling QE with intelligent agents
• Implementing multi-agent coordination
• Building CI/CD quality pipelines

PACT Principles

19-Agent Fleet

Coordination Patterns

Hierarchical: fleet-commander → [generators] → [executors] → quality-gate
Mesh: test-gen ↔ coverage ↔ quality (peer decisions)
Sequential: risk-analyzer → test-gen → executor → coverage → gate

Success Criteria

✅ 10x deployment frequency with same/better quality ✅ Coverage gaps detected in real-time ✅ Bugs caught pre-production ❌ Agents acting without human oversight on critical decisions ❌ Deploying all 19 agents at once (start with 1-2)

Core Concepts

QE Evolution

Core Premise: Agents amplify human expertise for 10x scale.

Key Capabilities

1. Intelligent Test Generation

// Agent analyzes code change, generates targeted tests
const tests = await qeTestGenerator.generate(prDiff);
// → Happy path, edge cases, error handling tests

2. Pattern Detection - Scan logs, find anomalies, correlate errors

3. Adaptive Strategy - Adjust test focus based on risk signals

4. Root Cause Analysis - Link failures to code changes, suggest fixes

Agent Coordination

Memory Namespaces

aqe/test-plan/*     - Test planning decisions
aqe/coverage/*      - Coverage analysis results
aqe/quality/*       - Quality metrics and gates
aqe/learning/*      - Patterns and Q-values
aqe/coordination/*  - Cross-agent state

Memory Operations (MCP Tools)

CRITICAL: Always use aqe memory store with persist: true for learnings.

1. Store data to persistent memory:

// Store test plan decisions (persisted to .agentic-qe/memory.db)
aqe memory store \
  --key "aqe/test-plan/pr-123" \
  --namespace "aqe/test-plan" \
  --value '{...}' \
  --json

2. Retrieve prior learnings before task:

// Query patterns before starting test generation
const priorData = await aqe memory get --key "aqe/learning/patterns/test-generation/*" --namespace "aqe/learning" --json

// Use patterns to guide current task
if (priorData.success) {
  console.log(`Loaded ${priorData.patterns.length} prior patterns`);
}

3. Store coverage analysis results:

aqe memory store \
  --key "aqe/coverage/auth-module" \
  --namespace "aqe/coverage" \
  --value '{...}' \
  --json

Three-Phase Memory Protocol

For coordinated multi-agent tasks, use the STATUS → PROGRESS → COMPLETE pattern:

// PHASE 1: STATUS - Task starting
aqe memory store \
  --key "aqe/coordination/task-123/status" \
  --namespace "aqe/coordination" \
  --value '{...}' \
  --json

// PHASE 2: PROGRESS - Intermediate updates
aqe memory store \
  --key "aqe/coordination/task-123/progress" \
  --namespace "aqe/coordination" \
  --value '{...}' \
  --json

// PHASE 3: COMPLETE - Task finished
aqe memory store \
  --key "aqe/coordination/task-123/complete" \
  --namespace "aqe/coordination" \
  --value '{...}' \
  --json

Blackboard Events

Example: PR Quality Pipeline

// 1. Risk analysis
const risks = await Task("Analyze PR", prDiff, "qe-regression-risk-analyzer");

// 2. Generate tests for risks
const tests = await Task("Generate tests", risks, "qe-test-generator");

// 3. Execute + analyze
const results = await Task("Run tests", tests, "qe-test-executor");
const coverage = await Task("Check coverage", results, "qe-coverage-analyzer");

// 4. Quality decision
const decision = await Task("Evaluate", {results, coverage}, "qe-quality-gate");
// → GO/NO-GO with rationale

Implementation Phases

Phase 1 Example

# Week 1: Deploy single agent
aqe agent spawn qe-test-generator

# Weeks 2-3: Generate tests for 10 PRs
# Track: bugs found, test quality, review time

# Week 4: Measure impact
aqe agent metrics qe-test-generator
# → Tests: 150, Bugs: 12, Time saved: 8h

Limitations & Strengths

Agents Excel At

• Volume: Scan thousands of logs in seconds
• Patterns: Find correlations humans miss
• Tireless: 24/7 testing and monitoring
• Speed: Instant code change analysis

Agents Need Humans For

• Business context and priorities
• Ethical judgment and trade-offs
• Creative exploration ("what if" scenarios)
• Domain expertise (healthcare, finance, legal)

Best Practices

Trust Progression

Month 1: Agent suggests → Human decides
Month 2: Agent acts → Human reviews after
Month 3: Agent autonomous on low-risk
Month 4: Agent handles critical with oversight

Agent Coordination Hints

coordination:
  topology: hierarchical
  commander: qe-fleet-commander
  memory_namespace: aqe/coordination
  blackboard_topic: qe-fleet

preload_skills:
  - agentic-quality-engineering  # Always (this skill)
  - risk-based-testing           # For prioritization
  - quality-metrics              # For measurement

agent_assignments:
  qe-test-generator: [api-testing-patterns, tdd-london-chicago]
  qe-coverage-analyzer: [quality-metrics, risk-based-testing]
  qe-security-scanner: [security-testing, risk-based-testing]
  qe-performance-tester: [performance-testing]

Related Skills

• holistic-testing-pact - PACT principles deep dive
• risk-based-testing - Prioritize agent focus
• quality-metrics - Measure agent effectiveness
• api-testing-patterns, security-testing, performance-testing - Specialized testing

Resources

• Agent definitions: .claude/agents/
• CLI: aqe agent --help
• Fleet status: aqe fleet status

Success Metric: Deploy 10x more frequently with same or better quality through intelligent agent collaboration.