Open-source "Taste Skill" for AI coding agents draws about 59,300 GitHub stars

An open-source set of instruction files called Taste Skill, designed to stop the low-quality, templated UI "slop" that AI coding agents tend to generate for frontends, has gathered about 59,300 stars on GitHub. Built by developer Leon Lin (Leonxlnx), the framework aims to equip major AI coding tools—Cursor, Claude Code, Codex, Gemini CLI, v0, Lovable and OpenCode among them—with rules that raise the quality of the UI they produce. According to the official site, an experimental latest release, v2, is recommended as the default; a single npx skills add command adds a SKILL.md to a project, which the agent then loads automatically.