What problem does it solve? UI reviews often produce vague, unsupported suggestions or accidental rewrites of a product's identity. This Skill audits one coherent product surface strictly against the design evidence that actually governs it, so every reported problem is backed by a contract, a runtime path, and a single deterministic correction. ## Core Features & Use Cases - Evidence-gated auditing: Traces routes, layouts, components, variants, and resolved tokens, then keeps only findings that pass contract, runtime, and correction proofs plus a falsification pass. - Read-only by design: Never modifies product source, installs dependencies, or commits; output is limited to a findings report and plan files under design-plans/. - Self-contained handoff plans: Writes one implementation plan per selected change using a structured template covering evidence chain, reuse, scope, validation, and stop conditions, so another agent can execute with zero prior context. - Use Case: Ask it to review your app's settings page for design-system drift; it returns a ranked table of at most three verified findings, then drafts an executor-ready plan for the one you pick. ## Quick Start Ask the agent to audit the settings page of the web app for design inconsistencies and prepare an improvement plan for the top finding.
