Agent skills that give coding agents design taste and browser check capabilities.
Agent skills for art direction, real archives, and automated browser checks.
Taste Skill is an agent plugin that prevents coding agents from generating generic landing pages. It forces them to adopt curated art directions, use real open archive photos, and evaluate their own UI in a browser.
It integrates with Claude Code and Codex via an agent skills CLI to guide front-end generation through a multi-step design process. The agent deals a researched design direction, builds with credited public domain images, and self-corrects based on real browser screenshots.
It suits developers and designers using AI coding agents who are tired of generic, uninspired UI outputs. It does not suit teams looking for traditional visual editors or those not using terminal-based coding agents.
Deals unique art directions derived from a study of hundreds of real websites.
Pulls real photography from sources like Wikimedia Commons and the Met with automatic credits.
Renders the output in a real browser across multiple widths to check and fix its own visual work.
Applies specific layout and data display logic for admin views, charts, and application screens.
Free under its open-source license, with no API keys or paid tiers required from the tool itself. The only cost is your agent token consumption and the hardware resources required to run Node and a browser.
880 open remote AI-training roles, updated November 15, 2026; most pay $20–$175 an hour. Zekai lists them free; the application goes to the employer’s platform.
It needs Node 22+ and an installed Chrome, Edge, or Chromium browser, plus internet access to fetch open archive pictures.
Yes, the design and review rules apply regardless of whether you use React, Vue, Svelte, or plain HTML.
They are sourced from open collections like Wikimedia Commons, Openverse, and the Met under public domain or permissive Creative Commons licenses.
See Zekai first in Google