WeChat Mini-Program Design
banjinbell/wechat-miniprogram-design
Protótipos interativos WeChat em frames iOS com handoff técnico e checklist CloudBase
Ver no GitHubSkill para designers e desenvolvedores criarem protótipos interativos de UI para mini-programs do WeChat em grade production. Gera previews lado-a-lado em frames iOS reais com a cápsula do WeChat, fluxos multi-tela e handoff pronto para implementação, incluindo integração com CloudBase e modelos de IA.
Instalação
npx convez-ai banjinbell/wechat-miniprogram-design1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.
Métricas
Casos de uso
- Design de mini-programs WeChat sem dependência de designers profissionais
- Geração rápida de 3 direções visuais distintas para validação de produto
- Handoff técnico com especificações CloudBase, modelos de IA e checklist preflight
- Prototipagem de fluxos multi-tela em canvas pan/zoom com interatividade ao vivo
Veredito

Solução muito especializada para o ecossistema WeChat, com excelente foco em evitar AI-slop e entregar handoff técnico real. Recomendado para times que desenvolvem mini-programs com integração CloudBase; fora desse contexto tem aplicabilidade limitada.
README
A tradução para o português deste README ainda está na fila.
WeChat Mini-Program Design
A coding-agent skill for designing production-grade WeChat mini-program UI — multi-screen interactive React prototypes inside a real iOS frame with the WeChat top capsule. Packaged as a Claude Code plugin; the core SKILL.md can also be read by other coding agents with filesystem access.
What This Does
Helps non-designers design WeChat mini-programs by showing them side-by-side device-frame previews instead of asking them to describe aesthetics in words. The deliverable is a clickable HTML prototype + a handoff.md your dev team can implement against — including AI feature wiring with the CloudBase preflight checklist.
Key Features
- Real iPhone Frame — Every screen lives inside an iOS bezel with the WeChat top capsule rendered correctly. No naked mockups.
- Flows, Not Screens — Default deliverable is a 4–6 screen flow on a pan/zoom canvas. The user reads transitions, not hero shots.
- Visual Style Discovery — 3 distinct home-screen previews side-by-side; pick the direction you like (or mix elements).
- Anti-AI-Slop — Avoids the "white card + system 苹方 + blue link + pink-purple gradient CTA" trap. 6 curated design systems with committed visual theses.
- CJK-First Typography — Real Chinese content, real product nouns, real dates. Latin display fonts pair to the CJK weight, not the other way.
- CloudBase-Ready Handoff — Generated
handoff.mdincludeswx.cloud.extend.AIcall shapes, the 5-step preflight checklist (envId, 成长计划, Token Credits, model group), and default model picks (deepseek-v3.2/hunyuan-image-v2).
Installation
Via Claude Code Plugin Marketplace
Two commands as two separate Claude Code messages (don't paste both at once):
/plugin marketplace add https://github.com/banjinbell/wechat-miniprogram-design
After that finishes:
/plugin install wechat-miniprogram-design@banjinbell
Then trigger with /wechat-miniprogram-design:wechat-miniprogram-design or just say "帮我设计一个日记小程序" — it auto-triggers.
Use the HTTPS URL. The shorter banjinbell/wechat-miniprogram-design form may make Claude Code try SSH, which can fail if GitHub isn't in your known_hosts.
Claude Code Manual Installation
git clone https://github.com/banjinbell/wechat-miniprogram-design.git
cp -r wechat-miniprogram-design/skills/wechat-miniprogram-design ~/.claude/skills/
Triggers as /wechat-miniprogram-design (no plugin namespace).
Other Coding Agents
Cursor, Cline, Aider, Codex, Gemini CLI, OpenCode — point the agent at this repo and ask it to use the skill:
https://github.com/banjinbell/wechat-miniprogram-design
The agent should start from skills/wechat-miniprogram-design/SKILL.md and load referenced support files on demand. The skill body uses capability-name language (e.g. "materialize a starter component") and defers shell-specific mechanisms to its built-in Runtime Adapter table, so it doesn't depend on Claude-Code-only tools.
Usage
Design a new mini-program
> "帮我设计一个日记小程序,五个屏幕:首页 / 写日记 / 加载中 / 漫画结果 / 分享卡"
The skill will:
- Ask 6 structured questions in one shot (product, user moment, flows, fidelity, brand, references)
- Generate 3 distinct home-screen previews in iOS frames
- Let you pick the direction (or mix elements across previews)
- Expand the chosen direction into all the screens you listed, on a single pan/zoom canvas
- Wire 2–4 live tweaks (color, type pairing, density, dark mode) so you can re-skin in-place
- Open the prototype in your browser + auto-generate
handoff.mdand a 2× PNG set
Refine an existing prototype
> "Phase 3 这版 hover state 太重了,能改细一点吗?还有把成长版的 mood face 换成更可爱的"
The skill detects Mode C (enhancement), reads every existing screen first, then applies modifications under explicit rules (no breaking capsule clearance, no breaking 44px tap targets, no overflow).
Recreate an existing mini-program
> "我想把这个小程序的设计抄一遍:https://github.com/some-team/their-miniprogram"
The skill switches to Mode B (reskin), imports tokens from the source repo, identifies the component library in use, and recreates the design at your canvas — pixel-faithful via lifted hex values rather than eyeballed screenshots.
Included Design Systems
Six pre-built, hand-tuned visual systems. The skill consults design-systems/README.md first, shortlists 2–3 candidates whose audience / mood / time-of-day / shareability / build complexity axes match your brief, then renders them as Phase 2 previews.
Dream Journal line
- Cosmic Nocturne — Late-night, navy & gold, serif-led, contemplative
- Liquid Iridescence — Soft, dreamy, oil-on-water gradients, modern editorial
- Paper Folio — Tactile paper texture, hand-bound book aesthetic, warm neutrals
Diary Comic line
- Manga Jump — 90s 日漫 spot color, screentone, kinetic, drop-shadow signatures
- Glass Dusk — Frosted glass plates, magic-hour gradients, premium consumer
- Y2K Bubble — Candy colors, chrome highlights, playful, share-first
Each system has a lightweight preview.md (~5KB, read during shortlisting) and a full design.md (~25KB, read only after the user picks the direction). Don't fit any of these? The skill will design a new system in the same shape and capture it for next time.
File Structure
wechat-miniprogram-design/ ← single-plugin marketplace
├── .claude-plugin/
│ ├── marketplace.json ← Claude Code marketplace catalog
│ └── plugin.json ← plugin manifest (rolling release)
├── skills/wechat-miniprogram-design/ ← the skill itself
│ ├── SKILL.md ← entry point
│ ├── Tweaks Panel.md ← sub-skill
│ ├── design-systems/ ← 6 design systems (selection index + previews + full specs)
│ ├── starters/ ← ios-frame, design-canvas, tweaks-panel, image-slot
│ └── references/ ← component-vocab, performance-budget, handoff-template
├── README.md
└── LICENSE ← MIT
License
MIT — see LICENSE. Use it, fork it, embed it into your own skill library; attribution welcome but not required.
Contributing
Issues and PRs welcome. The skill is opinionated by design — see What "Outstanding" Looks Like at the bottom of skills/wechat-miniprogram-design/SKILL.md for the bar it tries to clear.