Leon Lin

Leon Lin (@lexnlin), together with blueemi, built Taste Skill, an open-source framework aimed at preventing AI coding agents from generating generic, "slop" frontend interfaces — under the tagline "Less slop, designs pop."

The image-to-code skill listed on this platform solves a concrete problem: the gap between what a designer intends with a visual reference and what an AI agent actually builds from it. Rather than translating an image directly into code, the skill first generates its own design references, analyzes them in detail, and only then implements the frontend closely aligned to those references — a two-stage process that significantly improves visual fidelity to the original.

Taste Skill includes more than 13 individual skills, is compatible with common coding agents including Cursor, Claude Code, v0, and Lovable, and is backed by more than 20 GitHub sponsors along with support from Fluxion AI, Vercel, and Kimi. An experimental v2, according to the project, brings a substantial rewrite with stricter design protocols for 2026.