alchaincyf/huashu-design
Huashu Design · HTML-native design skill for Claude Code · Claude Code 里 HTML 原生的设计 skill · 高保真原型 / 幻灯片 / 动画 + 20 设计哲学 + 5 维评审 + MP4 导出 · Agent-agnostic
What it solves
Huashu Design is a design-automation skill for AI agents (like Claude Code, Cursor, and Codex) that allows users to generate professional, delivery-ready designs using only text prompts. It eliminates the need for manual GUI tools like Figma or After Effects by enabling agents to produce high-fidelity interactive prototypes, presentation decks, and motion graphics directly from a terminal.
How it works
The project provides a structured set of instructions (SKILL.md), assets, and scripts that turn an AI agent into a professional designer. It employs several core mechanisms:
- Brand Asset Protocol: A strict 5-step process where the agent searches for official brand guidelines, extracts actual hex colors from assets, and creates a brand specification to avoid guessing colors.
- Design Direction Advisor: When requirements are vague, the agent runs three parallel logic paths (randomized selection, real-world award-winning references, and top-studio philosophies) to present three distinct visual directions for the user to choose from.
- Junior Designer Workflow: The agent follows a professional iterative process: clarifying assumptions, using placeholders for early feedback, and then refining content and variations.
- Technical Toolchain: It uses a library of 60 native HTML style libraries and a custom motion engine (Stage + Sprite model) to render designs. It includes scripts to export these HTML outputs into MP4, GIF, PDF, and editable PPTX files.
Who it’s for
It is designed for users who prefer a chat-based, agentic workflow over traditional graphical user interfaces, specifically those needing quick, high-quality product announcement animations, clickable App prototypes, or professional presentation slides without manually designing them.
Highlights
- Diverse Deliverables: Generates clickable iOS/Web prototypes, editable PPTX slides, and high-frame-rate motion graphics with BGM.
- Agent-Agnostic: Compatible with any agent that supports the "skills" framework (e.g., Claude Code, Cursor, OpenClaw).
- Anti-AI Slop: Uses specific CSS techniques (like
text-wrap: prettyand OKLCH colors) to avoid generic AI-generated visual patterns. - Local Execution: The rendering and export pipeline runs 100% locally with no required API keys for core design functions.
Related
- Project
- Project
- Project
- Project
- Project