Framework scaffolder

1. Ask for the language, framework, BDD yes/no, CI, and reporting. Default: Playwright + TS, or pytest-playwright for Python teams.
2. Bootstrap commands:
   • Playwright TS: npm init playwright@latest | Python: pip install pytest-playwright && playwright install --with-deps
   • Selenium 4: pip install selenium pytest (Selenium Manager fetches drivers automatically, so no webdriver-manager is needed) | Java: org.seleniumhq.selenium:selenium-java + TestNG/JUnit 5
   • Cypress: npm i -D cypress && npx cypress open
   • WebdriverIO: npm init wdio@latest .
3. Layout:

tests/          specs, grouped by feature
pages/          page objects (locators + actions, no asserts)
fixtures/       setup/teardown, auth state, data factories
config/         env files: base URL, timeouts (no secrets)
utils/          api clients, helpers
reports/

4. Page objects: expose user intent (login(user)), not raw clicks. Return the next page or component. Keep assertions in tests.
5. Auth: log in once through the API or UI and reuse the state. Playwright: storageState in a setup project. Selenium: inject cookies.
6. BDD (only if business people read the specs): Cucumber (JS/Java) or Behave (pip install behave; features/*.feature + features/steps/; run behave). Keep steps thin, have them call page objects, and avoid UI wording in Gherkin.
7. Reporting with Allure:
   • pytest: pip install allure-pytest, then pytest --alluredir=allure-results, then allure serve allure-results
   • Playwright: npm i -D allure-playwright, then reporter ['allure-playwright']
   • Attach screenshots, traces and logs on failure.
8. Config: retries are 0 locally and 1–2 in CI, but report them. Run in parallel. Tag smoke vs regression (@smoke, -m smoke, --grep @smoke).
9. Deliver the scaffold as files plus a README with run commands, then offer a CI job (see ci-cd-and-test-data).