audit-test-failures

Diagnose the root cause of a test that fails intermittently, passes locally but fails in CI, or only fails when run alongside other tests.

What This Command Does

Launches the test-failure-analyzer agent, which scans test files for the concurrency and isolation patterns that produce non-deterministic failures, and reports each with a root cause rather than just a location.

This is the diagnostic counterpart to audit-testing. Reach for testing to survey overall suite quality; reach for test-failures when a specific test is already misbehaving and you need to know why.

What It Checks

1. Missing await confirmation – async work with no wait, so the test finishes before the callback fires
2. Missing @MainActor – UI tests touching main-actor-isolated state without isolation
3. Shared mutable state in a @Suite – parallel tests mutating the same collection
4. Task.sleep standing in for a wait – timing assumptions that hold on a fast machine and break in CI
5. Missing .serialized trait – tests that cannot safely run in parallel but are not marked
6. Test-generated crashes – force unwraps and precondition failures inside test code
7. Date comparisons in #expect – assertions that depend on wall-clock timing

Usage

/axiom:audit test-failures

Related

• test-failure-analyzer – The agent that powers this command
• audit-testing – Suite-wide quality survey rather than one failing test
• test-debugger – Closed-loop debugging that re-runs tests until they pass, once you know the cause
• run-tests – Run tests and parse the .xcresult for failures