What's Inside the nexu-io/open-design Claude Code Plugin

What is nexu-io/open-design?

nexu-io/open-design is one of the largest single plugins in Flowy's index: 163 skills, one MCP server (open-design), and one command (od-contribute), 165 components total, indexed from a public GitHub repository with 89,242 stars. Its own summary describes a local-first desktop app that turns a coding agent into a design engine, producing real files such as HTML, PDF, PPTX, and MP4, and bills itself as an open-source alternative to Anthropic's Claude Design. Flowy indexes this repository; it does not own, bundle, or host any part of it, so use the listing page linked at the top of this guide for the current install path.

Which skills generate images, video, and audio?

A dozen skills wrap fal.ai's hosted models: fal-generate for general image and video work, plus single-purpose ones such as fal-upscale, fal-restore, fal-image-edit, fal-video-edit, fal-train for LoRA fine-tuning, fal-3d, and fal-lip-sync. Five more do the same for Venice.ai: venice-image-generate, venice-image-edit, venice-audio-speech, venice-audio-music, and venice-video. Single-provider skills round out the rest: OpenAI's image API (imagegen), Google's Gemini image API (imagen), OpenAI's Sora video API (sora), OpenAI text-to-speech (speech), replicate, and pixelbin-media with more than 85 endpoints. image-enhancer sharpens existing screenshots, ecommerce-image-workflow turns real product photos into listing images rather than invented concepts, and algorithmic-art, shader-dev, and threejs cover generative art, GLSL shaders, and 3D scenes.

What handles frontend code, Figma handoff, and keeping design from looking generic?

Seven Figma skills cover one workflow: figma-use runs Figma Plugin API scripts and is a prerequisite for the rest, figma-generate-design and figma-implement-design move a design between Figma and code either way, and figma-generate-library, figma-create-design-system-rules, and figma-code-connect-components build, document, and wire up a design-system library. frontend-design, frontend-skill, and shadcn-ui build interfaces directly in code, stitch-loop runs a critique-adjust-ship cycle against a brief, and image-to-code-skill and web-clone rebuild an existing image or site. faq-page and login-flow each generate one screen type, and ui-ux-pro-max is catalog-only, without its full upstream templates.

A large cluster stops that output from reading as generic AI design: taste-skill, gpt-tasteskill, soft-skill, minimalist-skill, brutalist-skill, and stitch-skill each push one visual point of view; taste-skill-v1 exists only for backward compatibility. plan-design-review scores each design dimension and flags AI-slop signals, design-review audits and fixes shipped UI with atomic commits, review-animations covers motion code the same way, and impeccable-design-polish preps a finished artifact for a live pass. Reference material backs both: platform-design distills more than 300 rules from Apple, Material Design 3, and WCAG 2.2, apple-hig, wpds, and web-design-guidelines apply Apple's, WordPress's, and Vercel's guidelines, and color-expert covers OKLCH color science and palette generation.

What covers animation, decks, and documents?

Eight official GSAP skills split the library by concern: gsap-core, gsap-react and gsap-frameworks for React versus Vue and Svelte, gsap-scrolltrigger, gsap-timeline, gsap-utils, gsap-plugins, and gsap-performance. flutter-animating-apps covers motion outside the web, and emil-design-eng and emilkowalski-motion bring Emil Kowalski's micro-interaction judgment to the work.

Presentation output splits between engines and fixed templates: pptx, pptx-generator, slides, and nanobanana-ppt generate a deck from a brief, ppt-keynote targets one visual language, and pptx-html-fidelity-audit checks an export against its source HTML. Several more, including deck-guizang-editorial and digits-fintech-swiss-template, are each locked to one named palette rather than general-purpose. Documents follow the same pattern: doc, docx, minimax-docx, pdf, and minimax-pdf create and edit Word files and PDFs, while resume-modern, release-notes-one-pager, data-report, and article-magazine each produce one document type.

What rounds out marketing, social templates, dev tooling, and the MCP server?

Marketing skills cover copy and creative work: copywriting and ad-creative write it, marketing-psychology frames it, competitive-ads-extractor and brand-extract pull material from ad libraries and a live website, and brandkit, paywall-upgrade-cro, and domain-name-brainstormer cover brand boards, pricing screens, and naming. Six skills each recreate one app's UI as a card: card-xiaohongshu, card-twitter, social-x-post-card, social-reddit-card, social-spotify-card, and frame-macos-notification. youtube-clipper, video-downloader, remotion, and video-hyperframes cover source and programmatic video, plus smaller tools like gif-sticker-maker and hatch-pet. A dev-ops corner covers agent-browser for browser automation, full-page-screenshot and screenshot for capture, and export-download-debugging and pr-feedback-quality-gate for broken exports and review comments.

The plugin's one MCP server, open-design, carries no description beyond its name, so treat it as the connection into the desktop app rather than a documented tool list. od-contribute appears as both a skill and a command: it walks a contributor, coder or not, through one of four paths (ship a skill or design system, translate documentation, fix a typo or write a post, or report a bug) and opens the pull request or issue for them.

Who should install this?

At 163 skills, treat this like a library rather than one tool: pick whichever skill matches the job in front of you, a Figma handoff, a GSAP animation, a Reddit-style card, rather than expecting to use all of it at once. Check the listing page linked at the top of this guide for the current install path.