bdd-playwright-pom

Write the feature file first in business-readable Given/When/Then language, then create matching step definitions, then add or extend POM methods the steps call. Step files should call POM methods instead of raw page locators whenever a suitable POM exists.

Move duplicated locator logic into POM methods on the existing page class, then update the tests or steps to call those methods. If no relevant POM exists for the page, the Skill pauses and asks for permission before creating a new POM file.

Use Vitest for unit tests and view, controller, or service coverage, and Supertest for Express HTTP route and app integration behavior. Reserve Cucumber plus Playwright for browser-driven end-to-end flows that genuinely benefit from BDD.

Prefer getByRole first, then getByLabel, then getByTestId, and avoid XPath unless no robust semantic locator exists. Use row-scoped or container-scoped assertions to prevent strict mode collisions and strict URL expectations to avoid false positives.

Avoid browser automation for pure API tests with no page interaction, non-testing infrastructure work, and scenarios that can be covered cleanly without a browser. Do not introduce a second test framework when the repo already has a clear local convention.