e2e-testing-patterns – skill Lập trình cho agent AI

E2E Testing Patterns

Build reliable, fast, and maintainable end-to-end test suites that provide confidence to ship code quickly and catch regressions before users do.

Use this skill when

• Implementing end-to-end test automation
• Debugging flaky or unreliable tests
• Testing critical user workflows
• Setting up CI/CD test pipelines
• Testing across multiple browsers
• Validating accessibility requirements
• Testing responsive designs
• Establishing E2E testing standards

Do not use this skill when

• You only need unit or integration tests
• The environment cannot support stable UI automation
• You cannot provision safe test accounts or data

Instructions

1. Identify critical user journeys and success criteria.

2. Build stable selectors and test data strategies.

3. Implement isolated tests with observable assertions and tracing; diagnose retries rather than counting a retry as an ordinary pass.

4. Run in CI with parallelization and artifact capture.

Safety

• Avoid running destructive tests against production.
• Use dedicated test data and scrub sensitive output.

Resources

• resources/implementation-playbook.md for detailed E2E patterns and templates.

Worked example

Input: invalid login sometimes appears successful because the test reads the error before rendering finishes. Use a dedicated fixture account and assert await expect(page.getByRole('alert')).toContainText('Invalid credentials'). Run the test without retries and confirm the dashboard remains inaccessible. Expected: the failure state is observed reliably; a delayed error cannot silently pass.

Inputs and prerequisites

An authorized test URL, isolated accounts/data, the installed browser runner and known success/failure states. The playbook uses project-specific routes and adapters; install only the dependencies your existing suite needs.

Limitations

• A mocked backend or payment provider proves only the mocked boundary; retain separate real integration checks.
• Automated accessibility scans miss interaction and assistive-technology problems.
• Retries, larger timeouts and updated snapshots can hide regressions; preserve first-failure evidence.
• Browser tooling cannot validate a locked or unavailable interactive environment. Report that gap and continue independent tests.