/analyze-flaky

Reads test results and identifies flaky tests — tests that pass sometimes and fail sometimes without any code changes.

Usage

/analyze-flaky [<results-file>]

All command combinations

Every --tool and --id option in one place → Command Combinations

Examples

# Analyze the most recent run results
/analyze-flaky

# Analyze a specific results file
/analyze-flaky reports/results-TEST22.json

What it does

1. Reads test results (pass/fail/error per test)
2. Classifies each failure:
   • Timing — element not ready, race condition
   • Selector drift — element moved between renders
   • Data dependency — test depends on external state
   • Environment — flaky only in CI, not locally
   • Real bug — consistently failing, not flaky
3. Reports flaky tests with confidence score and recommended fix
4. Suggests hardening changes (better waits, data isolation, etc.)

Output

A flaky analysis report with:

• Which tests are flaky (and their flaky rate)
• Root cause classification
• Recommended fix per test
• Tests with 0% flaky rate that are solid (good to know too)

After analyzing

• For selector/timing issues → run /heal-tests to auto-fix
• For data dependency issues → review test isolation in your fixtures
• For environment issues → check CI vs local environment differences