Code Investigate Test Failure

When a unit test is failing and you need to systematically find the root cause from test result XML and log output.

The full operational procedure is in How do I investigate a test failure?. The skill exists to surface that recipe under a name an LLM session can invoke, and to add the framing decisions a recipe cannot make on its own.

1. Ask which suite if the user hasn't specified — single suite like ores.accounts.tests, or rat for everything.
2. Ask whether a specific test or all failures in the suite are in scope. A single failure usually means deep dive; a sweep means catalogue them, then prioritise.
3. Run How do I investigate a test failure? end-to-end: enable logging → run suite → parse results → walk the investigation order (exception → test source → per-test log → suite log → re-run with trace → targeted debug logging).
4. When you find the cause, decide:
   • Test bug → fix the test; commit on a feature branch.
   • Production bug → fix the code, and keep the test that surfaced it (or add a tighter one); open a task in the current sprint.
   • Flaky test → file a task with the seed (RNG Seed from the result XML) so it can be reproduced.
5. Disable logging when you're done — keeps the rest of the test suite fast.

• How do I investigate a test failure? — the full procedure.
• How do I enable test logging? — the prerequisite.
• How do I run the tests? — the surrounding run loop.

• Unit test conventions — file structure, naming, helpers — to orient yourself once you open the failing test file.
• CMake setup — preset / output-directory background.
• unit-test-writer — sibling skill for authoring tests.