
Give AI a design sense. 28 open-source Agent Skills for interface art direction: 1 master skill that
Give AI a design sense. 28 open-source Agent Skills for interface art direction: 1 master skill that picks the right visual style, 27 specialists that execute it with production-quality guidance and WCAG 2.2 AA guardrails. Plain markdown, MIT.
·
Give AI a design sense.
Install · Skill catalog · How it works · Verification · Showcase · Docs · Maintainers
Ochrelle is an open-source ecosystem of Agent Skills for design intelligence: one master skill that reasons about which visual style a product needs, and 27 specialist skills that execute a specific style with production-quality guidance.
The name is coined, not borrowed. Ochrelle derives from ochre, the earth pigment people have used to make surfaces meaningful for tens of thousands of years, with a made-up suffix chosen so the whole word is unique: at naming time it had zero exact matches on the web and npm. The repository itself is plain markdown, CSS, and Python that you can read, copy, and audit in an afternoon; the closest thing to an installer is a small copy script (scripts/install.sh).
| License | MIT (see LICENSE) |
| Format | Agent Skills: a SKILL.md with YAML frontmatter plus supporting references/ files per skill |
| Status | 2.0.0 (see CHANGELOG.md) |
Ask an AI coding agent to "make it look good" and you will usually get the same interface back: a centered card, a soft gradient, border-radius: 12px, Inter, and a shadow somewhere between Tailwind's shadow-md and shadow-lg. The result is technically correct and aesthetically generic, because three ingredients are missing:
Ochrelle packages these ingredients as skills an agent can load on demand. The agent keeps its own coding ability; Ochrelle supplies the taste, the rules, and the checklists.
The public, documented install path is Claude Code, which loads Agent Skills from a skills directory.
# 1. Get the skills
git clone https://github.com/izynu/ochrelle.git
cd ochrelle
# 2. Copy whole skill directories into your Claude Code skills folder
mkdir -p ~/.claude/skills
cp -r skills/ochrelle-design-master ~/.claude/skills/
cp -r skills/ochrelle-minimalism ~/.claude/skills/
# 3. Start a new Claude Code session and ask for design workInstalling on any agent: the repository ships scripts/install.sh, a dependency-free installer that covers the two common cases. ./scripts/install.sh --claude copies all 28 skills into ~/.claude/skills, and ./scripts/install.sh <destination> copies them into any folder, such as a project's skills/ directory. Add --only <skill-name> to install a single skill (for example ./scripts/install.sh --claude --only ochrelle-minimalism). The script only writes to the destination; the full compatibility matrix for Claude, Cursor, Codex, Gemini CLI, Aider, and other agents lives in docs/installation.md.
Notes:
SKILL.md.ochrelle-design-master is the orchestrator. It interviews the product context (audience, platform, brand temperature, content density), compares candidate styles, and picks one, or a primary with an accent. It then delegates to the matching specialist. Use it whenever the style is not already decided.
| Skill | Type | Best for |
|---|---|---|
ochrelle-skeuomorphism | Specialist | Real-world metaphors: calculators, note apps, tape decks, vintage controls, tactile familiarity |
ochrelle-neumorphism | Specialist | Soft-UI panels on monochrome surfaces: music players, smart-home controls, device dashboards |
ochrelle-glassmorphism | Specialist | Frosted translucent layers over vivid media: overlays, players, widgets, OS-like panels |
ochrelle-claymorphism | Specialist | Playful, friendly products: kids' apps, onboarding flows, brand moments with chunky soft 3D |
ochrelle-brutalism | Specialist | Statement sites, portfolios, counter-design, raw typographic honesty |
ochrelle-minimalism | Specialist | Productivity tools, docs, enterprise dashboards, content-first reading experiences |
ochrelle-maximalism | Specialist | Editorial, culture, fashion, and campaign pages that celebrate abundance |
ochrelle-liquid-glass | Specialist | Modern OS-level translucency: adaptive, dynamic material layers à la visionOS/iOS |
ochrelle-bento-grid | Specialist | Modular card compositions: landing pages, feature summaries, information-dense dashboards |
ochrelle-spatial-ui | Specialist | Depth-driven and 3D interfaces: AR/VR, visionOS-style windows, parallax scenes |
ochrelle-flat-design | Specialist | Solid-color, no-shadow production skins: SaaS dashboards, admin panels, settings screens, docs |
ochrelle-material-design | Specialist | Android and Google-ecosystem products: elevation levels, ripples, Material You dynamic color |
ochrelle-neo-brutalism | Specialist | Loud indie brands: thick black borders, hard offset shadows, sticker badges for merch, events, and creator tools |
ochrelle-y2k | Specialist | Dot-com futurism: candy plastics, chrome gradients, and gel buttons for rave, fashion, and revival campaigns |
ochrelle-cyberpunk | Specialist | Neon-noir atmosphere: games, club nights, security and developer brands, dark portfolios |
ochrelle-retro-design | Specialist | Mid-century Americana: cafes, diners, breweries, vinyl stores, food packaging, event posters |
ochrelle-editorial-design | Specialist | Magazine craft on the web: blogs, digital magazines, essays, press pages, story-driven marketing |
ochrelle-memphis-design | Specialist | Postmodern play: squiggles, terrazzo, and clashing pastels for youth, music, and party brands |
ochrelle-japandi | Specialist | Quiet warmth: wellness and mindfulness apps, boutique hospitality, calm productivity, skincare and tea brands |
ochrelle-frutiger-aero | Specialist | 2004-2013 glossy optimism: aqua gel, glass shine, and nature-tech for throwback marketing and kiosks |
ochrelle-corporate-memphis | Specialist | Flat vector illustration systems: onboarding, HR tools, empty states, community pages, friendly SaaS heroes |
ochrelle-vaporwave | Specialist | 80s/90s consumer-tech nostalgia: Roman busts, VHS glitch, and sunset gradients for music releases and streetwear |
ochrelle-pixel-art | Specialist | 8-bit/16-bit sprite discipline: games, launchers, emulation frontends, retro-gaming brands, achievement systems |
ochrelle-terminal-console | Specialist | Text-mode UI: monospace grids, box-drawing frames, phosphor palettes for developer tools, log viewers, and status pages |
ochrelle-hand-drawn | Specialist | Sketchbook warmth: education platforms, kids' tools, creative apps, onboarding and how-to guides |
ochrelle-art-deco | Specialist | Geometric luxury: boutique hotels, cocktail bars, theaters, jewelry and spirits brands, gala invitations |
ochrelle-futuristic-hud | Specialist | Sci-fi instrument panels: robotics and drone telemetry, launch events, mission-control-style pages |
Every skill directory is self-contained: a SKILL.md (frontmatter + workflow) and at least a references/style-guide.md with concrete tokens, component recipes, anti-patterns, and WCAG 2.2 AA notes.
A typical agent session flows through three hops:
SKILL.md's frontmatter: the description tells it when the skill is relevant, without burning context on the full body.ochrelle-design-master loaded, the agent interviews your product context (audience, platform, brand temperature, content density), scores the candidate styles, and commits to a direction, or you skip this step by loading a specialist directly.references/ files are read on demand: the style guide for tokens and rules, implementation examples for complete code, the quality checklist for a final pass. This progressive-disclosure pattern keeps the expensive content out of context until it is actually needed.Because every skill is plain markdown, the same files work as reading material for humans. Reviewing a skill in a PR is exactly like reviewing any other doc.
ochrelle-glassmorphism/
├── SKILL.md
│ ---
│ name: ochrelle-glassmorphism # must equal the directory name
│ description: Frosted translucent ... # what + when, 1-1024 chars
│ license: MIT
│ metadata:
│ version: "2.0.0"
│ ---
│
│ # Glassmorphism
│ (when to use, the workflow, pointers into references/)
└── references/
├── style-guide.md # required for specialists: tokens, rules,
│ # component recipes, WCAG 2.2 AA notes
├── implementation-examples.md
└── quality-checklist.mdThe frontmatter follows the Agent Skills specification: name (1-64 chars, ^[a-z0-9]+(-[a-z0-9]+)*$, identical to the directory name), description (1-1024 chars), plus optional license, compatibility, metadata, and allowed-tools. Everything else (reference resolution, placeholder scanning, the specialist style-guide requirement) is enforced by the bundled validators, so drift gets caught in CI instead of in front of your users.
A summary matrix of the ten original specialist styles; the full dimension-by-dimension comparison across all 27 specialists lives in docs/style-comparison.md.
| Style | Signature cue | Surface and depth | Contrast risk | Reach for it when |
|---|---|---|---|---|
| Skeuomorphism | Physical metaphor | Textured, beveled, "real" materials | Low (high fidelity aids recognition) | The UI should feel like a familiar object |
| Neumorphism | Dual light/dark shadow | Monochrome, extruded/pressed | High (low-contrast edges) | Monochrome control panels, calm hardware UIs |
| Glassmorphism | Frosted blur | Translucent layers over vivid backgrounds | Medium (depends on backdrop) | Media-forward overlays and widgets |
| Claymorphism | Puffy 3D softness | Chunky radius, stacked soft shadows | Low to medium | Playful, friendly, youth-oriented products |
| Brutalism | Raw HTML honesty | Hard borders, default-ish typography | Low (usually extreme contrast) | Statement design, portfolios, subcultures |
| Minimalism | Subtraction | Flat or near-flat, whitespace-led | Low | Productivity, reading, enterprise tools |
| Maximalism | Controlled overload | Layered pattern, rich color, big type | Medium | Editorial, culture, fashion, campaigns |
| Liquid Glass | Refractive adaptive material | Dynamic translucent "glass" that reacts to content | Medium to high | Modern OS-like surfaces, dynamic content over media |
| Bento Grid | Modular tiling | Card grid, mixed spans, rounded corners | Low | Feature summaries, dashboards, landing pages |
| Spatial UI | Depth and perspective | Layered z-space, eye/hand affordances | Medium | AR/VR, 3D scenes, immersive product demos |
If you want a one-line decision procedure before reading any comparison:
ochrelle-skeuomorphismochrelle-neumorphismochrelle-glassmorphism (or ochrelle-liquid-glass for a modern, adaptive OS feel)ochrelle-claymorphismochrelle-brutalism (or ochrelle-maximalism for rich editorial abundance)ochrelle-minimalismochrelle-bento-gridochrelle-spatial-uiWhen in doubt, or between two of the above, load ochrelle-design-master and let it route.
Example 1, you know the style:
Load the
ochrelle-minimalismskill and redesign this settings page. Follow its style guide and keep the result at WCAG 2.2 AA contrast.
Example 2, you don't know the style:
I'm building a smart-home dashboard for a hardware startup. Use
ochrelle-design-masterto choose betweenochrelle-neumorphismandochrelle-liquid-glass, explain the trade-off, then apply the winning style.
Example 3, restructuring a page:
Use the
ochrelle-bento-gridskill to restructure this landing page into a bento composition: hero tile, three feature tiles, one stats tile. Keep copy from the current page.
Example 4, checking your own work:
Using the
ochrelle-glassmorphismskill's anti-patterns list, review this component and list every violation with a fix.
The repository ships two stdlib-only validators and a unittest suite. CI runs all three on every push and pull request to main. Run them yourself, expect the numbers below:
# 1. Lint every skill directory (frontmatter, references, placeholders)
python3 scripts/validate_skills.py
# expect: 28/28 skill(s) passed
# 2. Check repository structure (docs, showcase, workflow, links, empty files)
python3 scripts/validate_repository.py
# expect: 9/9 check(s) passed
# 3. Run the unit tests (includes negative tests against fixtures)
python3 -m unittest discover -s tests -v
# expect: 29 tests, all passingBoth validators support --json for machine-readable output; validate_repository.py also accepts --root and validate_skills.py accepts --skills-dir for use outside a checkout.
examples/web-showcase/ is a static, dependency-free page that renders the same card component in the ten original specialist styles side by side. It is also the source of the showcase linked from the main website.
# Option A: just open it
open examples/web-showcase/index.html # macOS
xdg-open examples/web-showcase/index.html # Linux
# Option B: serve it (recommended, avoids local-file quirks)
python3 -m http.server 8000 --directory examples/web-showcase
# then visit http://localhost:8000The showcase is a hand-built illustration, not a generated artifact. See Known limitations.
ochrelle/
├── skills/ # 28 Agent Skills: 1 master + 27 specialists
│ └── <skill-name>/
│ ├── SKILL.md # YAML frontmatter + workflow body
│ └── references/ # 3 files per specialist (style-guide.md,
│ # implementation-examples.md, quality-checklist.md);
│ # the master ships 7 reference files
├── examples/
│ └── web-showcase/ # static side-by-side demo (index/styles/script)
├── website/ # this project's website (Next.js, static export)
├── docs/ # architecture, installation, research, assets
├── scripts/ # stdlib-only validators + universal installer
│ ├── validate_skills.py
│ ├── validate_repository.py
│ └── install.sh
├── tests/ # unittest suite + validator fixtures
├── DEPLOY.md # publish the website on GitHub Pages
└── .github/ # CI + Pages workflows, PR & issue templatesThe project website (the page with the skill catalog, verification commands, privacy policy, and terms) lives in website/ and is a static Next.js export: no database, no server runtime.
.github/workflows/deploy-pages.yml builds and publishes it to GitHub Pages on every push to main, so the site is hosted for free at https://izynu.github.io/ochrelle/ once Pages is enabled. The complete walkthrough, including serving the site from your own domain, is in DEPLOY.md.
| Document | What it covers |
|---|---|
| docs/architecture.md | How the master/specialist split works and how skills compose |
| docs/installation.md | Per-platform install notes beyond the Claude Code quickstart |
| docs/style-comparison.md | The full style comparison matrix with per-dimension detail |
| docs/research-methodology.md | How the style guidance was researched and curated |
| docs/source-catalog.md | The source catalog behind the methodology |
| docs/examples.md | Worked prompt/response examples for every skill |
| docs/troubleshooting.md | Common failure modes and fixes |
Contributions are welcome: new style guidance, sharper anti-patterns, better accessibility notes, fixed references. Start with CONTRIBUTING.md, which covers the frontmatter spec, the quality bar (original content, WCAG 2.2 AA integration, complete code examples), and the PR checklist. If you are an AI agent working in this repository, read AGENTS.md first: it contains the rules your changes must satisfy, and the validators will enforce them.
Run the validators before opening a PR: the CI workflow (.github/workflows/validate.yml) runs exactly these commands, and a red build is an annoying way to learn about a broken reference.
Do skills execute anything? No. Skills are markdown documents (plus reference files). The agent reads them; nothing is installed, compiled, or run. The only executable code in this repository is the stdlib-only Python validator in scripts/ and the scripts/install.sh copy script, and neither runs unless you (or CI) invoke them.
Can I use this outside Claude Code? Any agent that can read files and follows the Agent Skills format can use these skills. The mechanics differ per platform; see docs/installation.md.
Can I use Ochrelle commercially? Yes, MIT. See LICENSE.
Can I edit a skill for my own team? Yes. Fork the directory, tweak the tokens and rules to your brand, and keep the directory self-contained so the validators still pass. If your improvement is generic, upstream it via CONTRIBUTING.md.
Why markdown instead of a database or API? Skills must travel with the agent, be reviewable in PRs, and work offline. Markdown with a strict frontmatter header is the lowest-common-denominator that satisfies all three, and it is exactly what the Agent Skills ecosystem expects.
How do I add a twenty-ninth style? Open a Skill proposal issue first (naming and routing need to fit the master skill), then follow CONTRIBUTING.md. The validators and tests will walk you through everything else.
Being honest up front, because design advice given with false confidence is worse than none:
Ochrelle is maintained by Zynu (@izynu) and Shraddha (@xhraddha). The copyright line in LICENSE, the citation authors in CITATION.cff, and the contact channels in SECURITY.md, SUPPORT.md, and CODE_OF_CONDUCT.md all point at the repository's GitHub channels. If ownership ever moves, update those files and the repository URLs in CHANGELOG.md and the docs.
MIT. Skills are documentation and code samples , use them freely, including commercially, with attribution appreciated but not required beyond the license notice.
This repo does not document an install step we could read. Check the README on GitHub.
Paste this prompt into your agent. It reads this page and does the setup for you.
Read https://aiagentslisting.com/mcp/ochrelle to learn what the "Ochrelle" MCP server does and how to install it. Add it to my coding agent's MCP configuration as documented on that page, then confirm the server connects and list the tools it exposes.Agents can also browse this directory over MCP at https://aiagentslisting.com/api/mcp
This server doesn't publish a machine-readable tool manifest, and it doesn't expose a public endpoint we could ask. Connect the server locally to see what it exposes.
We could not run this server: it ships no npx package for our sandbox and no public endpoint. The four checks that need a live server are left out, and the grade tops out at B until one exists.
Maintain this server? Claiming proves you own the listing and earns the last 4 points.
Claim this listingPaste this line near the top of your repository's README. The badge always renders the current score.
[](https://aiagentslisting.com/mcp/ochrelle)Nothing comparable is listed yet.
No tagged releases on record.
Connect to the AI Agents Listing MCP server at https://aiagentslisting.com/api/mcp using the HTTP transport and complete the sign-in when it asks. Then call the submit_review tool for the listing "ochrelle" with a rating from 1 to 5 and a short note on what worked and what did not.No reviews yet
Be the first to share your experience with this tool.
One email a week. New agents, MCP servers and skills, and what is actually getting traction.