Best Claude Skills for Design (2026)

Most "best skills" lists are copied README summaries. This one isn't: we installed all 214 design skills in our catalog, ran each against a baseline and scored four criteria. 91 passed cleanly, average score 8.8/10. The one to start with: Codex GPT Image.

Ranked by SkillProof Score · how we score

  1. 1

    Codex GPT Image

    Generates gpt-image-2 images via local Codex OAuth, no OPENAI_API_KEY needed.

    From our test: Ran the bundled CLI for real, end to end: auth-status found a live Codex session, then `generate` actually hit the Codex Images backend and wrote a genuine 1.85MB PNG of a red bicycle on a white…

    Tested · Works ★ 10.0/10 by ningzimu

  2. 2

    Slide Wright

    Proposes an invented theme and a real 2-slide preview before it ever builds the full deck.

    From our test: Built the same 'Q3 sales results' request two ways and rendered both in a real browser: a direct rewrite produced one plain black-on-white bullet slide, while following the skill produced an actual…

    Tested · Works ★ 10.0/10 by arifszn

  3. 3

    Compose App Icon

    uv-bundled CLI pair that authors and JSON-Schema-validates Apple Icon Composer .icon packages

    From our test: Ran the bundled 25-test pytest suite for real (25/25 passed), then fed the same icon document through create_icon.py two ways: a naive version ignoring the SKILL.md 'Gotchas' section failed schema…

    Tested · Works ★ 10.0/10 by giginet

  4. 4

    HTML Explainer

    Builds polished, source-grounded, playable single-page or slide-deck web explainers with a real quality loop.

    From our test: Real A/B, rendered in-browser: the skill build produced a genuinely playable Doppler-shift canvas demo with color-coded compressed/stretched wavefronts, editorial serif/sans typography, and a real…

    Tested · Works ★ 10.0/10 by ds-vibe

  5. 5

    Frontend Design

    Makes Claude's UI output look designed instead of generated.

    From our test: Same landing-page brief run with and without the skill: the with-skill version had a real type scale, intentional palette and none of the neon-gradient tells. Night and day.

    Tested · Works ★ 9.6/10 by Anthropic

  6. 6

    WCAG 2.1 AA Web UI Audit

    Deterministic WCAG 2.1 AA audit -- keyboard, contrast, zoom/reflow, forms/ARIA, status messages -- ending in a severity-rated remediation backlog.

    From our test: Ran the bundled `run_axe_playwright.js` for real with no playwright/axe installed: it printed install guidance and exited 0 instead of crashing, exactly as documented. On a 3-line login-form snippet…

    Tested · Works ★ 9.6/10 by dvcrn

  7. 7

    Frontend Slides

    Builds zero-dependency, fixed-16:9 HTML decks or converts PPTX, avoiding AI-slop defaults.

    From our test: Built the same pitch deck twice: the plain version fell into purple-gradient-on-white with system fonts, the guided one used a real font pairing and a scaling stage that rendered with no overflow in…

    Tested · Works ★ 9.6/10 by zarazhangrui

  8. 8

    Preline Theme Generator

    Generates full light/dark Preline CSS theme files from a hue or brand hex via a bundled local Node script.

    From our test: Ran the actual bundled script with zero setup and no network calls, producing a 682-line CSS file with a full OKLCH scale and dark-mode chart tokens, versus a 16-property hand-written sheet with none.

    Tested · Works ★ 9.6/10 by htmlstreamofficial

  9. 9

    Algorithmic Art

    Builds seeded p5.js generative art with sliders in one branded, self-contained interactive HTML file.

    From our test: templates/viewer.html ships a fully-wired Anthropic-branded seed navigator (prev/next/random/jump, parameter sliders, PNG download) that the skill forces every sketch through, so the output is…

    Tested · Works ★ 9.6/10 by ThinkInAIXYZ

  10. 10

    AI-Firstify

    Scores a project against 9 AI-first design principles across 7 dimensions, then can fix what it finds.

    From our test: Ran a real partial audit on this very repo (Skill Agregator): baseline gave a vague paragraph, the skill-guided run flagged concrete, verifiable facts (no CLAUDE.md, no .claude/skills/ dir, but .env…

    Tested · Works ★ 9.6/10 by techwolf-ai

How this ranking works

Each skill gets a fresh install in Claude Code, three trigger phrasings, and the same task run with and without it. Scores: installation (5), trigger reliability (5), output vs baseline (10), documentation honesty (5) — normalized to a 10-point SkillProof Score. Full rubric on the methodology page. Queue-verdict skills sit at the bottom until testing completes.

New here? Start with how to install Claude skills — or grab the whole setup in one command with a curated pack ($10).

On a different AI tool? These tested instructions adapt: see our conversion guides for Cursor rules, Copilot instructions, Windsurf and Custom GPTs.