QE Defect Intelligence

Purpose

Guide the use of v3's defect intelligence capabilities including ML-based defect prediction, pattern recognition from historical data, and automated root cause analysis.

Activation

• When predicting defect-prone code
• When analyzing failure patterns
• When performing root cause analysis
• When learning from past defects
• When prioritizing testing based on risk

Quick Start

# Predict defects in changed code
aqe defect predict --changes HEAD~5..HEAD

# Analyze failure patterns
aqe defect patterns --period 90d --min-occurrences 3

# Root cause analysis
aqe defect rca --failure "test/auth.test.ts:45"

# Learn from resolved defects
aqe defect learn --source jira --status resolved

Agent Workflow

// Defect prediction
Task("Predict defect-prone code", `
  Analyze PR #456 changes and predict defect likelihood:
  - Historical defect correlation
  - Code complexity factors
  - Author experience with module
  - Test coverage gaps
  Flag high-risk changes requiring extra review.
`, "qe-defect-predictor")

// Root cause analysis
Task("Analyze test failure", `
  Investigate recurring failure in AuthService tests:
  - Collect failure history (last 30 days)
  - Identify common patterns
  - Trace to potential root causes
  - Suggest fixes using 5-whys analysis
`, "qe-root-cause-analyzer")

Prediction Models

1. Change-Based Prediction

await defectPredictor.predictFromChanges({
  changes: prChanges,
  factors: {
    codeChurn: { weight: 0.2 },
    complexity: { weight: 0.25 },
    authorExperience: { weight: 0.15 },
    fileHistory: { weight: 0.2 },
    testCoverage: { weight: 0.2 }
  },
  threshold: {
    high: 0.7,
    medium: 0.4,
    low: 0.2
  }
});

2. Pattern Learning

await patternLearner.learnPatterns({
  source: {
    defects: 'jira:project=MYAPP&type=bug',
    commits: 'git:last-6-months',
    tests: 'test-results:last-1000-runs'
  },
  patterns: [
    'code-smell-to-defect',
    'change-coupling',
    'test-gap-correlation',
    'complexity-defect-density'
  ],
  output: {
    rules: true,
    visualizations: true,
    recommendations: true
  }
});

3. Root Cause Analysis

await rootCauseAnalyzer.analyze({
  failure: testFailure,
  methods: [
    'five-whys',
    'fishbone-diagram',
    'fault-tree',
    'change-impact'
  ],
  context: {
    recentChanges: true,
    environmentDiff: true,
    dependencyChanges: true,
    similarFailures: true
  }
});

Defect Prediction Report

interface DefectPrediction {
  file: string;
  riskScore: number;  // 0-1
  riskLevel: 'critical' | 'high' | 'medium' | 'low';
  factors: {
    name: string;
    contribution: number;
    details: string;
  }[];
  historicalDefects: {
    count: number;
    recent: Defect[];
    patterns: string[];
  };
  recommendations: {
    action: string;
    priority: string;
    expectedRiskReduction: number;
  }[];
}

Pattern Categories

Pattern	Detection	Prevention
Null pointer	Static analysis	Null checks, Optional
Race condition	Concurrency analysis	Locks, atomic ops
Memory leak	Heap analysis	Resource cleanup
Off-by-one	Boundary analysis	Loop invariants
Injection	Taint analysis	Input validation

Root Cause Templates

root_cause_analysis:
  five_whys:
    max_depth: 5
    prompt_template: "Why did {effect} happen?"

  fishbone:
    categories:
      - people
      - process
      - tools
      - environment
      - materials
      - measurement

  fault_tree:
    top_event: "Test Failure"
    gate_types: [AND, OR, NOT]
    basic_events: true

Integration with Issue Tracking

await defectIntelligence.syncWithTracker({
  source: 'jira',
  project: 'MYAPP',
  sync: {
    defectData: 'bidirectional',
    predictions: 'create-tasks',
    patterns: 'update-labels'
  },
  automation: {
    flagHighRisk: true,
    suggestAssignee: true,
    linkRelated: true
  }
});

Coordination

Primary Agents: qe-defect-predictor, qe-pattern-learner, qe-root-cause-analyzer Coordinator: qe-defect-intelligence-coordinator Related Skills: qe-coverage-analysis, qe-quality-assessment