Side-by-side model comparison with rubric scoring.
Scan the product manifest — ontology types, functions, actions, policies, and dataset shells.