TESTING THE HARNESS ITSELF

PRACTICE LAB CONNECTION In the Resume Review Agent lab (Chapters 68 and 69), you will wire the eval suite into a continuous-check step, so that running the checks becomes automatic whenever the lab code changes. Test the prompts below against that step and its results. ARTIFICIAL INTELLIGENCE ASSISTANT PROMPTS [START OF PROMPT] Write unit, integration, and regression test cases for this component: [PASTE LAB OR PRODUCTION COMPONENT CODE HERE, e.g. permission resolver]. Cover at minimum one correctness case and one adversarial/edge case per test type, and flag any existing behavior the component relies on that is untested. Output format: test code or pseudocode grouped by test type, plus a short list of untested behaviors found. [END OF PROMPT] [START OF PROMPT] Design a continuous-evaluation gate for our release pipeline: specify exactly what eval suite runs on every proposed change, the score threshold that blocks a merge, and how to handle a borderline drop that has a defensible cause (such as fixing an unsafe shortcut). Current eval suite and recent score history: [PASTE LAB OR PRODUCTION EVAL RESULTS HERE]. Output format: a gate specification (trigger, suite, threshold, override process). [END OF PROMPT]