creative-first-ui
creative-first-ui 是一个为AI编码代理设计的视觉优先网页设计技能。它能将AI代理生成的网站从传统的文本+图标布局转变为以AI生成图像、视频和3D资产为核心的沉浸式、滚动驱动体验。该技能提供19个创意指导部分和14种视觉模式,帮助AI代理产出具有电影级转场和获奖级质量的独特网站设计,适用于Claude Code、Cursor、Codex、Copilot等多种AI编码工具。
npx skills add https://github.com/yasserstudio/creative-first-uiBefore / After 效果对比
1 组使用前,AI代理通常生成基于通用图库照片、图标网格和卡片布局的模板化网站,缺乏独特性和视觉冲击力。
使用后,AI代理能够产出以定制AI生成英雄视觉、滚动驱动叙事和电影级转场为特色的网站,达到获奖级设计质量。
description SKILL.md
creative-first-ui
Visual-first web design skill for AI coding agents. Build scroll-driven, immersive websites where AI-generated imagery, video, and 3D assets ARE the design — not decoration.
Works with Claude Code, Cursor, Codex, Copilot, Antigravity, and other AI coding agents.
What is this?
An AI agent skill that transforms how your coding agent designs websites. Instead of generating generic text+icon layouts, it produces visual-first, scroll-driven experiences built around AI-generated imagery.
Before this skill: Stock photos, icon grids, card layouts, template-looking pages. After this skill: Custom AI-generated hero visuals, scroll-driven storytelling, cinematic transitions, award-winning-quality output.
Install
npx skills add https://github.com/yasserstudio/creative-first-ui
Features
Design Philosophy
- 19 sections of creative direction, from ideation to pre-delivery checklist
- Visual-first approach — the image/video/3D IS the design, text serves it
- Creative ideation framework — visual metaphors, concept pairing, anti-obvious checks, mood grids
- Hero design rules — custom visuals, surgical gradient overlays, stats tickers, entrance animations
Visual Patterns (14)
- Full-bleed video backgrounds
- Scroll-driven frame sequences (Apple-style)
- 3D objects as hero (WebGPU/Three.js/Spline)
- Editorial image layouts
- Exploded views
- Visual crossfades
- Clip-path reveals
- Visual story sequences
- Parallax layer swaps
- Morphing visuals (scroll zoom)
- Video scrubbing
- Horizontal scroll
- Split-screen reveals
- Text masking (image through letterforms)
AI Asset Generation
- Direct generation via inference.sh CLI — 50+ image models, 40+ video models
- Tool comparison — Midjourney vs Flux vs Gemini vs Kling vs Runway
- Prompt templates per industry
- Asset pipeline — FFmpeg, Sharp, gltf-transform compression commands
Advanced Techniques
- CSS scroll-driven animations (
animation-timeline: scroll()) - Shader effects (Shaders.com, Three.js EffectComposer)
- Noise/grain textures (SVG feTurbulence)
- View Transitions API
- Rive animations
- Kinetic typography
- Magnetic/fluid cursor effects
- Bento grid layouts
- Sound design (opt-in)
Industry-Specific Design (13 Industries)
Automotive, Consumer Tech, SaaS, Fashion & Luxury, Health & Wellness, Health & Food Tracking, Architecture, Food & Beverage, Travel, Gaming, Finance, Education, Creative Agency — each with hero strategy, color direction, typography, patterns, and AI prompt templates.
Framework Integration
- React / Next.js — GSAP hooks, ScrollTrigger cleanup, next/image, React Three Fiber, View Transitions
- Astro — islands architecture, View Transitions, GSAP/Lenis lifecycle, content collections
Production Quality
- Apple HIG alignment (semantic color tokens, 8px spacing, 44px touch targets)
- Vercel Web Interface Guidelines alignment
- Performance budgets, iOS/Safari gotchas, loading states
- WCAG AA accessibility, reduced motion, high contrast
- Conversion optimization (CTA design, trust signals, visual psychology)
- 25-item pre-delivery checklist
File Structure
creative-first-ui/
SKILL.md # Main skill (1,500+ lines, 19 sections)
references/
examples.md # Award-winning sites by pattern & industry
snippets.md # Ready-to-use code (2,200+ lines)
industries.md # Design direction for 13 industries
react-nextjs.md # React/Next.js integration
astro.md # Astro integration
asset-pipeline.md # Image/video/3D optimization
CHANGELOG.md # 15 versions documented
ROADMAP.md
ENGINEERING.md
README.md
The Workflow
1. Tell the agent what you're building
2. It asks: "Can you generate AI visuals?"
3. Creative ideation → 3 hero concepts (safe → bold)
4. Generate assets (via infsh CLI or your preferred tool)
5. Agent builds the page around the visuals
6. Iterate: screenshot → identify issues → fix → repeat
Keywords
ui-design ux-design web-design landing-page scroll-animation gsap three-js webgl ai-generated-art visual-design creative-coding frontend hero-section parallax scroll-driven-animations midjourney kling inference-sh react-three-fiber astro nextjs tailwind accessibility performance conversion-optimization
License
MIT - Yasser's Studio
forum用户评价 (0)
发表评价
暂无评价
统计数据
用户评分
为此 Skill 评分