fix

This skill diagnoses and fixes failing or flaky Playwright tests using a systematic taxonomy. It addresses the frustrating problem of tests that fail intermittently or pass locally but fail in CI, giving a repeatable process to find the root cause instead of guessing.

The workflow reproduces the failure (including burn-in runs with `--repeat-each` and parallel-worker runs to expose flakiness), captures a full trace with `--trace=on`, and then categorizes the failure into one of four buckets using an included decision-tree taxonomy: timing/async, test isolation, environment, or infrastructure. For each category it prescribes targeted fixes, such as replacing `waitForTimeout` with web-first assertions and adding missing awaits, removing shared mutable state and resetting per-test data, matching viewport and timezone between local and CI, and tuning retries, workers, and timeouts for CI runners. It verifies the fix by running the test ten times and requiring all to pass, and finishes with prevention recommendations (CI retries, trace on first retry, documenting the pattern). The bundled taxonomy file includes concrete before/after TypeScript examples.

It targets developers and QA engineers who maintain Playwright suites and need a disciplined debugging method for flaky tests. It runs local `npx playwright test` commands to reproduce, trace, and verify, which are safe developer-driven test executions, and edits test code to apply fixes.