Playwright Web UI Automation (workflow)

Operating mode (token-efficient)

• •Treat this skill as a router + governor.
• •Do not load multiple procedures. Select exactly one procedure below and follow it end-to-end.
• •Prefer small, reviewable changes; keep tests deterministic.

Routing (pick one procedure)

Shared non-negotiables (apply to all procedures)

1. •

   Stable selectors

   • •Prefer data-testid / data-test attributes or role-based locators.
   • •Avoid brittle CSS/XPath tied to layout.
2. •

   No fixed sleeps

   • •Do not use arbitrary waitForTimeout as a primary strategy.
   • •Use Playwright auto-wait + explicit assertions (expect(...)) with timeouts.
3. •

   Test isolation

   • •Each test must be runnable independently.
   • •Use API-level setup for test data when feasible; avoid shared mutable state.
4. •

   Artifact contract (for CI + triage)

   • •Standardize under: artifacts/playwright/
   • •Must include (at least on failure): trace, screenshot, and a machine-readable result (e.g., JUnit).
5. •

   No secrets in repo

   • •Credentials must come from CI secrets / local env vars.
   • •Never commit tokens, cookies, or real user passwords.

Minimal inputs you should capture before changing code

• •Target environment(s): local / dev / staging
• •Base URL and navigation entry point(s)
• •Auth model: none / basic / cookie / SSO (and how to bypass for tests)
• •Target browser matrix (Chromium-only vs cross-browser)
• •Test data strategy (seeded fixtures, dedicated test tenant)

Verification

• •

  If you changed skills:

  • •Prefer host-repo tooling if present:
    • •node .ai/scripts/lint-skills.cjs --strict
  • •Always run the local validator:
    • •node .ai/skills/testing/test-web-playwright/scripts/validate-skill.cjs
• •

  If you changed tests/config:

  • •npx playwright --version
  • •npx playwright test
  • •If reports are generated:
    • •npx playwright show-report artifacts/playwright/report (path may vary by config)

Boundaries

• •Do not edit .codex/skills/ or .claude/skills/ directly (generated).
• •Do not introduce new test frameworks if Playwright already exists.
• •Do not rely on production data or production credentials.
• •Do not disable assertions to "make tests pass"; fix the underlying determinism issue.

Reconnaissance-then-action workflow (borrowed)

When adding or debugging Playwright E2E tests:

1. •

   Reconnaissance

   • •Confirm the target app is running and reachable.
   • •Use Playwright's inspector or trace viewer to inspect the rendered DOM.
   • •Identify stable selectors (data-testid, data-test, or role-based locators).
   • •Prefer explicit assertions with auto-wait over arbitrary timeouts.
2. •

   Action

   • •Implement the interaction using stable selectors.
   • •Assert on outcomes (visible UI state, route, network behavior) rather than layout details.

If Cypress and Playwright skills are both loaded

• •Do not attempt to use both frameworks in the same test suite.
• •Choose the framework already present in the repo (or explicitly requested by the user) and proceed with that skill's procedures.