How the AI-native UX workflow connects A feature request passes through the UX skill into the IDE agent. The agent pulls structure from Figma, product meaning and evidence from Notion, and existing patterns from the codebase, then either ships a reversible detail, proposes options, or surfaces a blocker. Blockers return to Notion as research questions. Shipped work lands in the Cense interface. structure meaning + evidence what exists unresolved question returns to research Feature request “add the risk score” UX skill how the agent reasons Cursor agent in the IDE Pulls context before it writes code. Acts alone on reversible details only. OpenAI · reasoning model Figma components, variants, states, where they’re used Notion design rules · component contracts research insights · open questions Codebase what already exists, what can actually ship Ships it reversible detail Offers options product decision Names a blocker open question Cense wallet evidence a bank analyst can defend