---

name: mutation-testing description: | Verify test quality by injecting mutations into code and measuring catch rate. Calculate mutation score (0.00-1.00) showing test effectiveness. Auto-generate missing tests to improve coverage. Integrate with Serena for continuous mutation tracking. Use when: improving test quality, validating test effectiveness, generating missing test cases, measuring code coverage gaps.

skill-type: ANALYTICAL shannon-version: ">=5.6.0"

mcp-requirements: recommended: - name: serena purpose: Mutation tracking, historical scoring, correlation analysis - name: stryker purpose: Mutation testing framework integration - name: pitest purpose: Java/Kotlin mutation testing - name: mutmut purpose: Python mutation testing

allowed-tools: All

Mutation Testing - Quantified Test Quality

Purpose

Measure test effectiveness by injecting mutations (intentional bugs) into code and tracking how many your tests catch. Calculates mutation score (0.00-1.00) to quantify test coverage gaps. Auto-generates additional tests targeting uncaught mutations. Integrates with Serena MCP for continuous tracking and trend analysis.

When to Use

• Measuring test effectiveness beyond code coverage
• Identifying weak test areas (low mutation scores)
• Auto-generating tests for mutation-resistant code
• Validating test quality gates (require 0.80+ mutation score)
• Tracking mutation improvements over time
• Comparing mutation scores across teams/projects

Core Metrics

Mutation Score Calculation:

CopyScore = (Killed Mutations / Total Mutations) × 1.0
Range: 0.00 (no tests catch bugs) to 1.00 (all bugs caught)

Score Interpretation:

• 0.90+ Excellent test suite, catches most bugs
• 0.80-0.89 Good coverage, minor gaps
• 0.70-0.79 Acceptable but needs improvement
• <0.70 Poor coverage, significant blind spots

Workflow

Phase 1: Mutation Generation & Execution

1. Inject mutations: Stryker/PIT/mutmut inject bugs
2. Run tests: Execute full test suite
3. Track kills: Count caught vs escaped mutations
4. Calculate score: Derive 0.00-1.00 metric

Phase 2: Serena Integration

1. Push metrics: Send mutation_score, killed_count, escaped_count to Serena
2. Track history: Store scores by commit, branch, timestamp
3. Alert on regression: Flag if score drops >0.05
4. Trend analysis: Show mutation score trajectory

Serena Push Example:

Copy{
  "metric_type": "mutation_score",
  "project": "task-app",
  "value": 0.87,
  "components": {
    "auth": 0.92,
    "api": 0.85,
    "ui": 0.79
  },
  "killed": 156,
  "escaped": 23,
  "timestamp": "2025-11-20T10:30:00Z"
}

Phase 3: Gap Analysis & Test Generation

1. Identify mutations: List escaped mutations (bugs tests miss)
2. Pattern detection: Group by type (boundary, null, operator)
3. Generate tests: Auto-create test cases targeting gaps
4. Validate new tests: Confirm they catch previously escaped mutations

Auto-Generated Test Example:

Copy# Escaped mutation: changed > to >= in boundary check
# Auto-generated test catches this:
def test_score_boundary_at_90():
    """Catches mutations in >= vs > comparisons"""
    assert get_grade(90) == 'A'  # Catches >= → > mutation
    assert get_grade(89) == 'B'  # Catches >= → > at boundary

def test_null_mutations():
    """Catches null/None comparison mutations"""
    assert validate_email(None) == False  # Catches is None → == None
    assert validate_email("") == False    # Catches == "" mutations

Real-World Impact

E-Commerce Cart System:

• Initial mutation score: 0.71 (many edge cases uncaught)
• Generated 34 new tests targeting mutations
• Improved score to 0.88 (caught boundary bugs, null handling)
• Prevented 12 production bugs in checkout flow

Financial API:

• Mutation score: 0.82 initially
• Serena tracked regression: dropped to 0.79 (new code)
• Auto-generated 8 tests for numeric precision mutations
• Restored to 0.85, prevented rounding errors

Success Criteria

✅ Mutation score ≥0.80 for critical code ✅ No score regression >0.05 between commits ✅ Auto-generated tests all pass ✅ Serena tracking shows improvement trend ✅ Gap analysis identifies specific weak patterns