Soft assertions and evidence

Soft checks

• Expect.Soft(locator) and Expect.Soft(screen) record each failure and let the test go on. On NUnit the failures land on the test result, as inside Assert.EnterMultipleScope, and the test fails when it ends with every failure listed. On xUnit v3 they stay on the session, and the test fails when it is disposed, with every failure in one message.
• Expect.Soft(Browser) keeps the failures of ToHaveURLAsync, ToHaveTitleAsync, and ToHaveClassAsync the same way.
• Expect.Poll(read) retries a value. ToBeAsync and ToSatisfyAsync(predicate) work on both adapters. E2E.NUnit also adds ToMatchAsync with any NUnit constraint: await Expect.Poll(() => ReadCountAsync()).ToMatchAsync(Is.GreaterThan(3));. A read that throws, or a matcher that refuses the value’s type, is one failing sample, not the end of the poll.

Evidence screenshots

App.ScreenshotAsync(label) saves a redacted screenshot under the attempt’s screenshots/ directory and returns the absolute path of the file it wrote (<number>-<label>.png); the report records the path relative to the artifact root. It masks secure fields only, after a Secret fill too, so a secret the app shows anywhere else is in the image. The screenshots a model sees (Vision = On, Only, the agent’s own looks, e2e mcp, and e2e explore findings) work the same way: a generic secret filled into a plain input, or a value the app shows outside a secure field, reaches your model provider, so use test credentials. It fails with UNSUPPORTED_CAPABILITY only on an engine that keeps no screenshot artifact. The web, mobile, and Windows engines keep screenshot artifacts. A secret spelled into the label is redacted before it names the file.