Acceptance criteria in the CAB cover error and edge paths as well as the happy path, which is what makes them usable as a test basis rather than a summary.After a build, the testing agent opens the application in a browser and works through it. It writes its own scenarios, each a complete user journey starting from navigation, and every scenario runs in a fresh browser session so that one cannot leave state behind that affects another. Only the pass or fail summary returns to your conversation.A build runs at most five test rounds. That caps repeated attempts within one request; it is not a limit on how often you can ask for testing.
