Test Metrics Dashboard

Test Metrics Dashboard

Data & Analysis skill for querying test execution history, identifying trends, and surfacing actionable quality metrics.

Activation

/test-metrics-dashboard

Key Metrics

Test Health Metrics

Data Collection

# Export Jest results to JSON
npx jest --json --outputFile=test-results/$(date +%Y-%m-%d).json

# Parse results for dashboard
jq '{
  date: .startTime,
  total: .numTotalTests,
  passed: .numPassedTests,
  failed: .numFailedTests,
  duration_ms: (.testResults | map(.endTime - .startTime) | add),
  pass_rate: ((.numPassedTests / .numTotalTests) * 100),
  flaky: [.testResults[] | select(.numPendingTests > 0)] | length
}' test-results/$(date +%Y-%m-%d).json

Trend Analysis

# Compare last 5 runs
for f in $(ls -t test-results/*.json | head -5); do
  jq --arg file "$f" '{
    file: $file,
    pass_rate: ((.numPassedTests / .numTotalTests) * 100 | floor),
    duration_s: ((.testResults | map(.endTime - .startTime) | add) / 1000 | floor)
  }' "$f"
done

Top Failing Tests

# Find most frequently failing tests across runs
for f in test-results/*.json; do
  jq -r '.testResults[] | select(.numFailingTests > 0) | .testFilePath' "$f"
done | sort | uniq -c | sort -rn | head -10

Run History

Store dashboard data in ${CLAUDE_PLUGIN_DATA}/test-metrics.log:

2026-03-18|95.2|4.1|312|82.5|3

Format: date|pass_rate|flakiness_rate|duration_s|coverage_pct|failed_count

Read history for trend detection:

# Coverage trending down?
tail -5 "${CLAUDE_PLUGIN_DATA}/test-metrics.log" | awk -F'|' '{print $5}' | sort -n | head -1

Composition

Feeds into:

• `/qe-quality-assessment` — quality gate decisions based on metrics
• `/test-failure-investigator` — investigate top failing tests
• `/coverage-drop-investigator` — when coverage trends down

Gotchas

• Metrics without baselines are meaningless — establish baselines before tracking trends
• Flakiness rate is underreported — a test that fails 1/100 times still breaks CI weekly
• Duration trends upward over time as test count grows — set alerts on rate of increase, not absolute value
• Agent may report metrics from a single run as "trends" — need 5+ data points for meaningful trends