← ClaudeAtlas

wechat-miniprogram-designlisted

Create distinctive, production-grade WeChat mini-program UI/UX design — multi-screen interactive prototypes that live inside a real iOS frame with the WeChat top capsule. Trigger whenever the user touches mini-program visual direction — designing from scratch, mocking an existing app, evolving a single screen into a full clickable flow, redesigning, reskinning, or reviewing a prototype. Fires on 小程序界面 / 小程序设计稿 / 小程序原型 / 微信小程序 UI / 界面设计 / 视觉方向 / 改版 / 重新设计, and on English equivalents (mini-program UI, WeChat screen design, interactive prototype, screen mockup, visual redesign). Fires even when the user doesn't say "design" explicitly — e.g. "帮我做个登录页", "画几个屏", "这个页面再优化下", "make this screen feel less generic". Helps non-designers discover their aesthetic through side-by-side device-frame previews rather than abstract style words. If a request touches a mini-program visual decision, this skill runs.
banjinbell/wechat-miniprogram-design · ★ 0 · Web & Frontend · score 70
Install: claude install-skill banjinbell/wechat-miniprogram-design
# WeChat Mini-Program Design Create multi-screen, capsule-aware, finger-ergonomic mini-program UI that looks like a real product running on a real iPhone — not a marketing mock. ## The prototype is NOT the mini-program (read this first) This is the single most important framing for the skill. The AI burns the most tokens when it forgets it. You are producing an **HTML React prototype that looks like a running mini-program**, viewed in Chrome. It is for design review, not for shipping to WeChat. Engineering converts the design into wxml/wxss/js later — they are a downstream consumer, not your runtime. What this means for your decisions while generating: - **Use modern CSS freely in the prototype.** `clip-path`, `backdrop-filter`, `:has()`, `mask-image`, view transitions, container queries, scroll-driven animations — all work in the Chrome that renders the prototype. Do not refuse a design choice because "WeChat WebView doesn't support it." That is a handoff concern, not a generation concern. - **Wire AI features through a single `aiComplete(prompt)` async hook.** The prototype's React component declares one such hook at the top; the body just calls it. The hook has two runtimes: in the prototype (Chrome) it mocks with realistic timing or, when available, calls `window.claude.complete`; in the production mini-program it calls `wx.cloud.extend.AI` directly — **no cloud function proxy needed**, the AI extension is the transport. The UI never changes between the two; only th