Taste Skill
A third-party plugin from Leonxlnx with thirteen skills for making AI-built websites and apps look designed, not templated. Most write design code: a main anti-slop skill, brutalist, minimalist, and high-end styles, a motion-heavy one, and one for redesigning an existing project. Five of the thirteen are not code-design skills. Three (brandkit, imagegen-frontend-web, imagegen-frontend-mobile) only make images and never write code. image-to-code makes design images first, then builds the site from them. full-output-enforcement is a general rule that stops the model from cutting its output short. One more, stitch-design-taste, writes a DESIGN.md file for Google Stitch. We did not write it and have not reviewed it.
Third-party, by Leonxlnx
September 19, 2026
Install the plugin
Plugins install through Claude Code’s plugin marketplace. Two commands, no files to place by hand.
- 1
Add the marketplace
Run this once in Claude Code. It registers the marketplace this plugin installs from (taste-skill), so every plugin in it becomes installable.
/plugin marketplace add Leonxlnx/taste-skill - 2
Install the plugin
Installs this plugin and everything it bundles. Restart Claude Code afterwards so it picks the new skills up.
/plugin install taste-skill@taste-skill
Leonxlnx publishes this plugin from a marketplace called taste-skill. We
list it here because people look for ways to make AI-built interfaces less
generic. We did not write it and have not reviewed it.
The main skill, design-taste-frontend, reads your brief, picks a design
direction, and checks its own work before it ships. design-taste-frontend-v1
is the older version, kept for projects that depend on how it behaved. The
style skills give you a look to aim for: industrial-brutalist-ui,
minimalist-ui, high-end-visual-design, and gpt-taste (heavy on scroll animation). Use
redesign-existing-projects when you already have a site and want to
upgrade it. stitch-design-taste writes a DESIGN.md file for Google Stitch.
Some skills do something else. brandkit, imagegen-frontend-web, and
imagegen-frontend-mobile only make images: brand boards, website sections,
and app screens. They don’t write code. image-to-code makes design images
first, then builds the site to match them. full-output-enforcement tells the
model to write complete code instead of trimming it. It works with any task,
not just design.
If you want one distinctive-UI skill and not a set, look at frontend-design too.
The skills it ships
- brandkit
- industrial-brutalist-ui
- gpt-taste
- image-to-code
- imagegen-frontend-mobile
- imagegen-frontend-web
- minimalist-ui
- full-output-enforcement
- redesign-existing-projects
- high-end-visual-design
- stitch-design-taste
- design-taste-frontend-v1
- design-taste-frontend