Ask for a real workflow inside the chat, then inspect the model transcript, the generated ToonUI, and the new catalog-driven prompt layers behind it.