What problem does it solve? Frontend work by AI coding agents often ignores existing repository conventions, introduces unrequested dependencies, or ships changes that break accessibility, performance, or type safety. This Skill enforces a conservative, evidence-led workflow so frontend design, review, and implementation stay minimal, safe, and consistent with the codebase. ## Core Features & Use Cases - Six structured modes: design, optimize, fix, analyze, review, and plan, each with its own checklist and output contract covering component architecture, state ownership, rendering models, and trade-offs. - Deep reference library: on-demand guidance for accessibility (WCAG 2.2 AA), Core Web Vitals performance budgets, TanStack Query and Zustand state patterns, TypeScript strictness, testing with Vitest and Playwright, security (XSS, CSRF, CSP), and design systems. - Safety guardrails: explicit constraints against unrequested migrations, dependency additions, auth or CSP changes, and broad refactors, plus a validation gate before every response. - Use Case: Ask the agent to review a React PR; it returns severity-tagged (P1/P2/P3) findings across a11y, performance, type safety, and security with file/line evidence and concrete fixes, without editing code unless asked. ## Quick Start Ask the agent to review your React component or design a new frontend feature, for example: review this LoanForm component for accessibility, performance, and type-safety issues.