test-flakiness

To detect flaky tests in CI logs, you can parse JUnit XML and plain-text test logs to build per-test pass/fail histories across multiple runs, then compute fail-rate thresholds to identify intermittent failures. The Skill aggregates these histories to surface nondeterministic results.

Flaky tests in JUnit suites are commonly caused by timing issues, order dependency, random seeds, resource leaks, external state, or floating-point comparisons. The Skill classifies likely causes by analyzing pass/fail histories across repeated CI runs.

Yes, you can analyze GitHub Actions test results for intermittent failures by applying the Skill to repositories with repeated CI runs. It parses test-results directories and saved local logs to identify nondeterministic failures and classify their causes.

You should quarantine flaky tests instead of fixing them when you need to reduce noise in pipelines during the polish phase. The Skill recommends quarantine, fixes, or monitoring based on computed fail-rate thresholds and optionally appends quarantined tests to regression-suite.md.

Yes, test-flakiness works without external dependencies because it operates directly on existing CI run logs, JUnit XML files, and plain-text test outputs to compute fail-rate thresholds and generate a flakiness report.