What problem does it solve? Design teams lack a repeatable way to test a design system against real product scenarios: interviews are ad hoc, prototypes drift from component rules, and review feedback is lost or mixed with builder mistakes. This Skill runs the full design-test loop with strict provenance and separation of duties. ## Core Features & Use Cases - Structured designer interview: Runs a six-section interview covering users, domain model, status lifecycles, screens, edge states, and test goals, producing a TEST.md scenario file. - Rule-constrained prototype build: Builds a working prototype exclusively from Recursica components and tokens, then dispatches the Barb reviewer subagent to check every screen against the design system's published rules before the designer sees it. - Findings with provenance: Captures designer feedback into FINDINGS.md and self-evaluation into EVAL_REPORT.md, separating design rules from builder defects and package gaps. - Use Case: A design lead wants to know whether the component library covers a task-management app. The Skill interviews the designer, builds the app from library components, has it rule-reviewed, and returns structured findings plus a package-defect list. ## Quick Start Ask the agent to run a design test: interview me about the tool I want to prototype, build it from our design system components, and capture my review as findings.
