When an AI coding agent can modify both implementation and tests, a green test suite needs careful interpretation. The agent might fix an implementation defect—or change the test until the defect is no longer detected. In our C#/.NET project, we implemented…