What problem does it solve? LLM-generated frontend code tends to fall back on the same templated aesthetics: AI-purple gradients, centered heroes, three equal feature cards, and generic glassmorphism. This Skill forces the agent to read the brief first, infer the right design direction, and apply concrete anti-default rules so the output does not look machine-generated. ## Core Features & Use Cases - Brief Inference and Design Read: Analyzes page kind, vibe words, references, audience, and constraints, then declares a one-line design read before writing any code. - Three Configuration Dials: Tunes DESIGN_VARIANCE, MOTION_INTENSITY, and VISUAL_DENSITY with signal-based inference tables and per-use-case presets. - Design System Mapping: Routes briefs to official packages (Fluent, Material 3, Carbon, Polaris, GOV.UK, shadcn/ui, Tailwind v4) or honest native-CSS implementations for aesthetics like bento, brutalism, and glassmorphism. - Hard Layout and Pre-Flight Rules: Enforces hero viewport fit, eyebrow restraint, CTA contrast and wrap checks, palette locks, and section-layout diversity, with audit-first handling for redesigns. - Use Case: Ask for a premium consumer cookware landing page and receive a React/Tailwind build with a non-default palette, real image assets, and a passing pre-flight audit instead of the usual beige-and-brass template. ## Quick Start Use the taste-skill to design and build a landing page for my SaaS product that does not look like a generic AI template.
