Blog · Ranking · 2026-05-28

Best Claude Code skills for UX/UI design — 2026 ranking.

The Claude Code skill ecosystem has eight design-focused plugins worth installing in 2026. Star counts below are live from GitHub on 2026-05-28. Each card lists what the skill is great at, what it lacks, who should pick it, and how to install it. No marketing spin.

By the ux-skill team · 10 min read · MIT-licensed open source

How we ranked

Not by stars. Stars correlate with awareness, not fit. We ranked by scope of capability — how much of a real designer's day the skill covers — and broke ties with depth of catalog. The top three could each be #1 for a different user. We say who should pick which.

The skills below are: ui-ux-pro-max-skill, open-design, taste-skill, huashu-design, stitch-skills, ux-skill v2, hallmark, and material-3-skill. We left off skills with fewer than 500 stars and skills not directly focused on UI/UX design (general taste, vibe-coding helpers, etc.).

ui-ux-pro-max-skill

nextlevelbuilder/ui-ux-pro-max-skill · MIT

83,958GitHub stars

~600structured entries

~50design styles

161industries

"design intelligence for building professional UI/UX"

What it's great at

The catalog. A Python recommender that returns a design system (style + palette + type + industry rules) in one call. The five-parallel-search architecture is the cleanest implementation of "queryable design taste" in the ecosystem and the community has voted on it heavily.

What it lacks

No deterministic linter. No brand-specific output beyond category-level styles. Single-skill invocation rather than a chained pipeline. Limited to Claude Code primarily.

Pick if

You want the most-installed, best-supported UX skill on GitHub and your workflow is "ask for a system, generate from system" in Claude Code.

$ /plugin install ui-ux-pro-max@nextlevelbuilder

github.com/nextlevelbuilder/ui-ux-pro-max-skill

open-design

nexu-io/open-design · MIT

54,406GitHub stars

19skills

71design systems

11README locales

"19 Skills · 71 Design Systems · Sandboxed preview · HTML/PDF/PPTX/MP4 export"

What it's great at

Scope. 19 distinct skills covering generation, preview, and multi-format export. Sandboxed preview is a real differentiator — most skills generate without showing. The export to PDF / PPTX / MP4 makes it a true presentation pipeline. AGENTS.md architecture means it's also usable with Cursor and other AI tools.

What it lacks

No anti-AI-slop linter. The 71 design systems are flavors, not deeply enforced brand languages. Heavy install footprint (Node project with sandbox runtime).

Pick if

You need design output you'll present — slide decks, hi-fi prototypes, video walkthroughs — and you want a sandbox preview before committing.

$ /plugin install open-design@nexu-io

github.com/nexu-io/open-design

taste-skill

Leonxlnx/taste-skill · MIT

25,202GitHub stars

13SKILL variants

3dial mental model

"gives your AI good taste. stops boring slop."

What it's great at

Brand discipline. They own the "anti-slop" identity in the Claude skill space and earned it by being early and consistent. The 3-dial mental model (DESIGN_VARIANCE, MOTION_INTENSITY, VISUAL_DENSITY) is the clearest user-facing abstraction in the category. Research-backed prompt with citations.

What it lacks

The 13 variants are flavors of taste, not a pipeline. No structured manifests. No linter. No CI integration. Bash + markdown only.

Pick if

You want a fast taste primer that's earned its reputation, your workflow is single-shot, and you've decided which variant matches your project.

$ /plugin install taste-skill@Leonxlnx

github.com/Leonxlnx/taste-skill

huashu-design

alchaincyf/huashu-design · MIT

15,455GitHub stars

20design philosophies

5dim review

62KBsingle SKILL.md

"hi-fi prototypes, slides, animations, 20 design philosophies, 5-dim review, MP4 export"

What it's great at

Philosophy-first generation. The 20 named design schools (Bauhaus, Swiss, Memphis, neo-brutalism, etc.) give the model a coherent frame to generate inside, not a loose taste vector. The 5-dimension review (visual, structural, hierarchy, motion, copy) is a clear rubric.

What it lacks

Single-file SKILL.md (62KB) is hard to extend. The 5-dim rubric is human-scored, not deterministic. No structured manifests outside the prose. No cross-IDE.

Pick if

You're an enthusiast picking a design school for a project (Memphis revival, neo-brutalism, Swiss editorial, etc.) and want the philosophy baked in.

$ /plugin install huashu-design@alchaincyf

github.com/alchaincyf/huashu-design

stitch-skills

google-labs-code/stitch-skills · Apache-2.0

5,762GitHub stars

Googleaffiliation

MCPserver-backed

"Skills library wired to the Stitch design MCP server"

What it's great at

Trust signal. It's Google-affiliated (Google Labs Code org). The skill library is wired to Stitch's MCP server, so it can fetch live design system data rather than ship static manifests. Useful if you're already in the Google design tooling ecosystem.

What it lacks

Requires the Stitch MCP server (network dependency). Smaller catalog than ui-ux-pro-max or ux-skill. No anti-slop discipline. Apache-2.0 is a slightly heavier license than MIT for some users.

Pick if

You're using Stitch already, you want a Google-affiliated official-flavored option, and you're comfortable with an MCP-server dependency.

$ /plugin install stitch@google-labs-code

github.com/google-labs-code/stitch-skills

ux-skill v2

Laith0003/ux-skill · MIT

14GitHub stars

998structured entries

35linter rules

72brand specs

22slash commands

17IDEs

"Python engine + 11 JSON manifests + 22 slash commands + 5 sub-agents + 120-rule deterministic linter + 131 brand DESIGN.md specs."

What it's great at

Architecture and rigor. Same Python recommender pattern as ui-ux-pro-max, plus four nets new — a deterministic regex linter, 72 named brand DESIGN.md specs, 57 motion presets, 148-component catalog. Pipeline-shaped: discover → recommend → generate → lint → polish, with state files chaining commands.

What it lacks

Awareness. 14 stars vs the leader's 84,000. v2 is alpha (2.0.0-alpha.1 a few weeks ago). Solo-maintained. No community contributor base yet. If you need momentum, install the leaders.

Pick if

You want a CI-gateable deterministic linter, you're building inside a named brand language (Apple, Stripe, Linear, Figma, Tesla, BMW, Notion, Spotify…), or you're outside Claude Code in Cursor / Windsurf / Copilot / Gemini / Aider.

$ /plugin marketplace add Laith0003/ux-skill
$ /plugin install ux@ux-skill

github.com/Laith0003/ux-skill

hallmark

Nutlope/hallmark · MIT

2,164GitHub stars

3IDEs supported

Vercelhosted site

"Anti-AI-slop design skill for Claude Code, Cursor, Codex."

What it's great at

Distribution. Hassan El Mghari's reach gives it strong install momentum. Multi-IDE intent from day one — Claude Code, Cursor, Codex. Clean Vercel-hosted landing site, real product polish.

What it lacks

Prompt-based, no linter. The anti-slop posture is in the prompt, not enforced on output. Smaller manifest depth than the leaders. No discovery step or chained pipeline.

Pick if

You like Nutlope's other open-source work and want an anti-slop skill that works across Claude Code, Cursor, and Codex without setup.

$ /plugin install hallmark@Nutlope

github.com/Nutlope/hallmark

material-3-skill

hamen/material-3-skill · MIT

955GitHub stars

30+components

MD3audit pass

"30+ components, design tokens, theming, MD3 audit"

What it's great at

Single-system depth. Knows Material Design 3 thoroughly — tokens, components, theming rules, the MD3 audit. If you're committing to MD3 for an Android-first product or a Flutter app, this is the most specialized skill in the list.

What it lacks

Single-system scope. It's only useful if you've already chosen Material 3. No recommendation lane, no anti-slop, no brand language outside MD3.

Pick if

You've committed to Material 3 for a Flutter or Android product and want a deep, specialized skill for that exact system.

$ /plugin install material-3-skill@hamen

github.com/hamen/material-3-skill

How to choose — by problem, not by stars

The 2x2 below picks the right skill based on what you're actually solving for. If your axis is awareness, install the leaders. If your axis is depth or fit, look at where you land.

Quadrant 1

Generic catalog + community

You want depth in the recommender and the most-installed option. Install ui-ux-pro-max.

Quadrant 2

Presentation pipeline

You need to ship a deck, prototype, or video walkthrough out of the AI. Install open-design.

Quadrant 3

Anti-slop primer (single-shot)

You want a taste-loaded prompt for a one-off generation. Install taste-skill or hallmark.

Quadrant 4

CI-gated brand language

You're building inside a brand (Apple / Stripe / Linear / Figma / etc.) and need a deterministic linter that fails on AI fingerprints. Install ux-skill v2.

Can you install more than one?

Yes — and most working developers do.

ui-ux-pro-max gives you the strongest recommender. ux-skill v2 gives you the only deterministic linter. taste-skill gives you the cleanest one-paste taste primer. They compose: run the recommendation, generate inside the brand spec, lint the output, ship.

The category isn't winner-take-all. Skills are small, MIT-licensed, and load on demand. Install three or four; pick by the surface of work you're doing right now.

The honest position

About the ux-skill ranking at #6

If we were ranking by stars, ux-skill would be #8 of 8 — 14 stars is the smallest installed base in the list. We ranked by scope (1,182 entries, 22 commands, 5 sub-agents, 35 linter rules, 131 brand specs, 17 IDEs) and put it where the architecture says it lands. Read that as a self-aware bias.

The cards above are accurate on what each tool actually ships. Star counts will move (they're live as of 2026-05-28). Pick by the problem, not the badge.

The two trends we're watching in 2026

First: structured manifests over taste prompts. Four of the top six skills use structured data (JSON manifests, philosophy files, design system catalogs). Pure prompt-only skills are losing ground.

Second: cross-IDE distribution. The Claude Code skill marketplace is the launch surface, but the actual installs split across Cursor, Windsurf, Copilot, Gemini, Aider. Skills shipping pip / npm / multi-IDE installers (open-design, ux-skill, hallmark) will likely consolidate the top of the ranking in the next 12 months.

Install ux-skill v2

The deepest catalog. The only linter. The brand specs.

1,182 structured entries, 120 deterministic rules, 131 brand DESIGN.md files. MIT licensed, no telemetry, no account. Three install channels, 17 IDEs.

$ /plugin marketplace add Laith0003/ux-skill

$ /plugin install ux@ux-skill

— or —

$ pip install uxskill

— or —

$ npx uxskill@alpha init