#eval

Reads an exported Copilot Studio agent and generates a graded, runnable test suite (happy-path, paraphrase, disambiguation, negative, knowledge-grounding, multilingual, and safety cases) plus a regression set, ready to run in the free Copilot Studio test panel.