swiftui-design-skill
agent
Warn
Health Warn
- License — License: MIT
- Description — Repository has a description
- Active repo — Last push 0 days ago
- Low visibility — Only 7 GitHub stars
Code Pass
- Code scan — Scanned 1 files during light audit, no dangerous patterns found
Permissions Pass
- Permissions — No dangerous permissions requested
Purpose
This tool provides a structured set of guidelines, design principles, and evaluation frameworks to help AI coding agents generate high-quality, visually appealing SwiftUI interfaces that avoid common AI design flaws.
Security Assessment
Overall Risk: Low
The code scan found no dangerous patterns and the tool requests no dangerous permissions. It does not appear to access sensitive data, execute shell commands, or make external network requests. The repository is essentially a collection of prompt instructions and Markdown documentation rather than executable software. No hardcoded secrets or malicious payloads were detected.
Quality Assessment
Quality is decent but community trust is currently minimal. The project is actively maintained, with the most recent push occurring today. It is properly licensed under the permissive MIT license. However, it suffers from very low visibility, having only 7 GitHub stars. This means the tool has not been widely peer-reviewed or battle-tested by a large developer community. The provided documentation and README are exceptionally thorough, detailed, and clearly explain the tool's mechanics.
Verdict
Safe to use — it poses no security risks and serves as a highly detailed, agent-agnostic prompt library, though its low community adoption means the actual design outputs depend entirely on your AI agent's capabilities.
This tool provides a structured set of guidelines, design principles, and evaluation frameworks to help AI coding agents generate high-quality, visually appealing SwiftUI interfaces that avoid common AI design flaws.
Security Assessment
Overall Risk: Low
The code scan found no dangerous patterns and the tool requests no dangerous permissions. It does not appear to access sensitive data, execute shell commands, or make external network requests. The repository is essentially a collection of prompt instructions and Markdown documentation rather than executable software. No hardcoded secrets or malicious payloads were detected.
Quality Assessment
Quality is decent but community trust is currently minimal. The project is actively maintained, with the most recent push occurring today. It is properly licensed under the permissive MIT license. However, it suffers from very low visibility, having only 7 GitHub stars. This means the tool has not been widely peer-reviewed or battle-tested by a large developer community. The provided documentation and README are exceptionally thorough, detailed, and clearly explain the tool's mechanics.
Verdict
Safe to use — it poses no security risks and serves as a highly detailed, agent-agnostic prompt library, though its low community adoption means the actual design outputs depend entirely on your AI agent's capabilities.
SwiftUI 前端设计 skill — 反 AI Slop 六条铁律、设计方向顾问、品牌资产协议、五维评审。支持 Claude Code / Cursor / Codex / OpenCode 等全部 AI agent 平台。
README.md
在你的 agent 里打一句话,拿回一个不 AI slop 的 SwiftUI 界面。
3 到 30 分钟,你能设计出一个有签名感的 iOS 界面——不是「AI 做的还行」那种水平,是看起来像有品味的设计师做的。
反 AI slop 六条铁律、设计方向顾问、品牌资产协议、五维评审——全部内置。给 skill 你的品牌色板,它会读懂你的气质;什么都不给,内置的 5 种设计语汇也能兜底到不出丑。
跨 agent 通用——Claude Code、Cursor、Codex、OpenCode、Hermes 都能装。
装上就能用
npx skills add wholiver/swiftui-design-skill -g -y
能做什么
| 能力 | 交付物 | 典型耗时 |
|---|---|---|
| SwiftUI 界面设计 | 可编译的 SwiftUI 代码 · 设计系统 token · 布局模式 | 10–15 min |
| 设计方向顾问 | 5 流派 × N 种设计哲学 · 推荐 3 方向 · 视觉锚点描述 | 5 min |
| 品牌资产集成 | brand-spec.md · 色板 · 字体 · 间距系统 | 5–10 min |
| 反 AI Slop 审查 | 逐条检查 · 具体修复代码 · 替代方案 | 3–5 min |
| 五维评审 | 雷达图评分 · Keep/Fix/Quick Wins · 可操作修复清单 | 3 min |
| 布局模式库 | 9 种常见布局 · 可复制的 SwiftUI 代码 | 即时 |
| 动画设计 | spring/parallax/pull-to-refresh · 可编译代码 | 5–10 min |
核心机制
反 AI Slop 六条铁律
这是 skill 里最硬的一段规则。所有设计都必须通过这 6 条检查:
| # | 铁律 | ❌ 禁止 | ✅ 替代方案 |
|---|---|---|---|
| 1 | 起点从上下文开始 | 空白画布凭空捏造 | 先问设计系统/UI kit/品牌资产 |
| 2 | 初级设计师模式 | 等完美方案 | 灰色占位符 > 烂 SVG |
| 3 | 给变体不给终稿 | 一个「最终答案」 | 2–3 个差异化方向 |
| 4 | 占位符 > 烂实现 | AI 生成的剪贴画 | 干净的灰色占位 + 文字标注 |
| 5 | 系统优先,别填充 | 塞满每一个像素 | 每个元素必须证明自己存在的理由 |
| 6 | 禁止 AI slop 模式 | 紫蓝渐变、emoji 图标、圆角卡片+左边框 | 单一暖色强调色、SF Symbols、衬线标题 |
详见 references/anti-ai-slop.md。
设计方向顾问
当需求模糊时,skill 会从 5 大设计流派中推荐 3 个差异化方向:
| 流派 | 特征 | 代表风格 |
|---|---|---|
| 信息型 | 数据优先、图表密集 | Bloomberg Terminal |
| 编辑型 | 杂志排版、衬线字体、大量留白 | NYT、Medium |
| 表现型 | 大胆配色、不对称布局、动效前置 | Stripe、Linear |
| 功能型 | 密集工具感、等宽字体点缀 | Things、OmniFocus |
| 温暖极简 | 柔和中性色、圆角、微妙质感 | Notion、Craft |
品牌资产协议
涉及具体品牌时强制执行 5 步硬流程:
| 步骤 | 动作 | 目的 |
|---|---|---|
| 1 · 问 | 用户有 brand guidelines 吗? | 尊重已有资源 |
| 2 · 搜 | 搜索品牌官方页面 | 获取真实素材 |
| 3 · 下 | 下载实际资产文件 | PNG/SVG logo、字体 |
| 4 · 验 | 验证颜色匹配官方来源 | 核对 hex 值 |
| 5 · 写 | 生成 brand-spec.md | 记录完整设计系统 |
质量门槛:5 个真实品牌色、10 个设计 token、2 个字体家族、8pt 间距网格。
五维评审
每个设计交付前必须通过 5 个维度的评审:
| 维度 | 评分标准 | 最低分 |
|---|---|---|
| 🎯 哲学一致性 | 设计是否贯彻了选定的设计哲学 | ≥ 7/10 |
| 📐 视觉层级 | 信息优先级是否清晰 | ≥ 7/10 |
| 🔧 细节执行 | 间距、字体、颜色是否精确 | ≥ 7/10 |
| ⚡ 功能性 | 布局是否服务用户目标 | ≥ 7/10 |
| ✨ 原创性 | 是否有 1 个签名细节 | ≥ 7/10 |
文件结构
swiftui-design-skill/
├── SKILL.md ← 核心定义(278 行)
├── references/
│ ├── anti-ai-slop.md ← 反 AI Slop 详细规则(268 行)
│ ├── layout-patterns.md ← 9 种布局模式 + 可复制代码(265 行)
│ ├── typography-color.md ← 字体层级 + 色彩系统(172 行)
│ ├── design-review.md ← 五维评审流程(151 行)
│ └── swift-extensions.md ← Color/Font/Animation 扩展(373 行,必备)
└── templates/
└── brand-spec.md ← 品牌规范模板(77 行)
与 swiftui-expert-skill 的区别
| SwiftUI Design Skill | swiftui-expert-skill | |
|---|---|---|
| 焦点 | 视觉设计、美学、品牌感 | 代码质量、性能、正确性 |
| 问题 | 「看起来怎么样?」 | 「代码写得对不对?」 |
| 输出 | 设计方向、色板、布局、评审 | 代码审查、Instruments 分析、API 现代化 |
| 互补 | 先用 design 确定方向 | 再用 expert 确保代码质量 |
两个 skill 可以配合使用:Design 确定「做什么」,Expert 确保「怎么做对」。
License
MIT — 随便用,但请保留原作者署名。
Reviews (0)
Sign in to leave a review.
Leave a reviewNo results found