What problem does it solve? Recruiting real users for early design iteration is slow and expensive, so wireframes and prototypes often go untested until late-stage validation. This Skill runs PerceptUI-style synthetic user sessions where LLM evaluators are conditioned on the 8 personas from prior UX research, producing natural-language rationales that expose usability issues before human testing. ## Core Features & Use Cases - Persona-Conditioned Evaluation: One LLM evaluator per persona from the A1 research matrix, answering a fixed question battery in first person with grounded rationales. - Activity-Specific Batteries: Distinct protocols for A3 (taxonomy and wireframe review supporting card sorting) and A4 (seven real prototype screens covering access, exploration, governance, assistant, and export). - Documented Design Iteration: Findings are synthesized by severity and frequency across the 8 personas, requiring at least one documented design change with before/after evidence. - Use Case: Before running the A5 SUS test with real participants, pre-validate the high-fidelity prototype screens with the 8 synthetic evaluators, capture their rationales, fix the top consensus findings, and report the process as complementary pre-validation. ## Quick Start Run a synthetic pre-validation of the A4 prototype screens using the 8 personas from A1 and produce a findings report with at least one documented design iteration.
