fix-flaky-tests — Claude Code skill: fix flaky tests

What fix-flaky-tests does

Everything runs through the Tuist CLI. The skill pulls reliability metrics with tuist test case show <id-or-identifier> --json, lists flaky runs with tuist test case run list Module/Suite/TestCase --flaky --json, then reads the failure message, file path and line number from tuist test case run show <test-case-run-id> --json. It opens the source at that path and edits it.

With no test given, it discovers them itself: tuist test case list --flaky --json --page-size 50. Tests are grouped by suite, since several flaky cases in one suite often share a cause. It also checks whether failures share a test_run_id — those may come from a process crash rather than a bug in the test.

The skill carries a catalogue of usual suspects: missing waits and race conditions, shared state between tests, dependence on network, time or hardcoded paths, and ordering. Crashes are read from crash_report, including EXC_BREAKPOINT, EXC_BAD_ACCESS and EXC_CRASH. Fixes are meant to be the smallest change that removes the cause, with no refactoring around it. Verification is a repeated run through xcodebuild test with -test-iterations and -run-tests-until-failure; the skill suggests starting at 50–100 iterations for fast unit tests and 2–5 for slower integration ones. One caveat: a race or a CI-only timing issue may not reproduce locally, and then the skill works from logs and code analysis instead.