### About We should have basic E2E tests that ensure our AI Catalog code successfully executes AI agents. This will catch any errors/incompatibilities with Duo Workflow Service that would cause AI agents to fail. **Risks** Potentially Duo Workflow Service itself can be unstable, and so these tests may be flaky, or fail for a period of time. This may cause problems for pipelines that expect to be great, for example, before releases. ### Proposal Work with a Test Engineer: - Validate that we could add E2E tests with Duo Workflow Service, the service itself may sometimes be unavailable, and the tests would not unduly block GitLab productivity due to broken tests Add basic E2E tests: - Create an agent - Agent test works - Assert that certain tools were used successfully. - As agentic results are non-deterministic, we want to test that the run was successful, not its result. - Create a Flow - Flow test works - Assert that certain tools were used successfully. - As agentic results are non-deterministic, we want to test that the run was successful, not its result. Already implemented: * ~~Empty list~~ * ~~List~~ * ~~Create agent~~ * ~~Edit agent~~ * Delete agent * Test run agent