Leonxlnx/taste-skill

What it solves

It prevents AI agents from generating "slop"—generic, boilerplate-looking user interfaces. It provides a set of portable instructions (skills) that guide AI coding agents to produce premium frontends with professional layout, typography, motion, and spacing.

How it works

The project provides a collection of SKILL.md files that act as portable instruction sets for AI agents. These can be installed via a CLI (npx skills add) or pasted directly into conversations with agents like ChatGPT, Cursor, or Claude Code.

There are two types of skills:

• Implementation skills: These output actual code, using specific design languages (e.g., minimalist, brutalist, or high-end visual design) and adjustable "dials" for variance, motion intensity, and visual density.
• Image-generation skills: These provide art direction for image generators to create reference boards (web, mobile, or brand kits), which can then be fed back into a coding agent for implementation.

Who it’s for

Developers and designers using AI coding agents to build websites and mobile apps who want to avoid generic AI-generated aesthetics and achieve a high-end, professional look.

Highlights

• Adjustable Design Dials: Control layout experimentation, animation depth, and information density via numeric settings.
• Diverse Visual Styles: Includes specialized skills for minimalist, brutalist, and "soft" premium aesthetics.
• Image-to-Code Pipeline: Ability to generate visual references first and then implement them in code.
• Framework Agnostic: Works across major frameworks like React, Vue, and Svelte.