UI tests are the most brittle layer of the testing pyramid. Selector-based tests break when element IDs or class names change, when layouts shift, or when interaction flows are updated. Recording-based test tools capture UI interactions but produce tests tightly coupled to the current UI state. Any redesign requires re-recording or manual script updates. The maintenance burden becomes so high that teams either invest heavily in UI test maintenance or abandon UI test coverage entirely. Neither outcome is acceptable for production applications with complex user workflows that require validation from the user's perspective.

Skyramp's UI test generator works from full-stack traces rather than UI element selectors. The trace collection layer captures the complete interaction between a user and the application stack, including both the UI interactions (clicks, typing, navigation) and the underlying API calls they trigger. The UI test generator processes these traces and produces tests that validate the complete user flow, including both the frontend behavior and the backend API responses it depends on. Because the tests are generated from real user behavior rather than scripted from specifications, they reflect actual usage patterns rather than theoretical workflows. The trace records the exact sequence of interactions, making generated tests accurate reproductions of real user journeys.

The output is a structured UI test that validates the complete user journey captured in the trace. The test covers the sequence of user interactions (clicks, text input, navigation steps), the API calls triggered at each step, the expected responses at each service boundary, and the final UI state at the end of the flow. Generated tests are deterministic: given the same trace input, the generator produces the same test every time. Tests can be run in Skyramp's isolated execution environment, which provisions all required service dependencies before the test run begins, eliminating failures caused by missing or misconfigured dependencies.

UI tests validate user-facing workflows end to end. API tests validate service behavior in isolation. Both are necessary for complete coverage. UI tests are slower and more complex but validate that the full stack works correctly from the user's perspective, catching integration issues that API tests miss. API tests are faster and more stable but cannot validate the user-facing behavior that emerges from the frontend-backend interaction. A complete test strategy uses API-level tests for rapid feedback on service behavior and UI-level tests for workflow validation from the user's perspective.