leonxlnx/taste-skill
Taste Skill is an agent skill repository containing `design-taste-frontend`, designed to guide AI coding agents in producing distinctive web and application layouts rather than generic templates. Intended for developers utilizing autonomous coding tools such as Claude Code and Codex, the system operates based on empirical studies of 247 curated websites. Its workflow follows six steps: analyzing the brief, dealing three of twenty researched visual directions for the agent to select from, establishing a unifying design concept, embedding credited media from open public archives, evaluating layout rendering across three viewport widths in a real browser, and iteratively fixing visual flaws over up to three review cycles. Version 2.1 also includes dedicated workflows for constructing application screens and dashboards.
The GitHub snapshot shows the repository is not archived and records an MIT license. Verified operational prerequisites require Node.js 22 or higher, an installation of Chrome, Edge, or Chromium, and network connectivity to retrieve open-archive media, without needing API keys. Installation is executed using `npx skills add https://github.com/Leonxlnx/taste-skill --skill "design-taste-frontend"`. For Claude Code, setup runs via `/plugin marketplace add Leonxlnx/taste-skill` followed by `/plugin install taste-skill@taste-skill`. For Codex, installation runs via `codex plugin marketplace add Leonxlnx/taste-skill` and `codex plugin add taste-skill@taste-skill`. Symlinking from a local clone is supported via `./install.sh`. Former companion skills have moved to separate repositories. For another design skill, compare Impeccable; its scope and workflow differ.
Data as of : 108973 stars · 7339 forks.