Token导航 LogoToken导航TokenDH.com
前端设计external-servicegithub未标认证来源可访问许可证需确认审计通过

stitch-ui-design-variants缝合 ui 设计变体

Agent Skill

用于辅助界面设计、视觉规范、排版、配色、布局和交互体验优化。它适合让 Agent 根据产品场景整理页面结构、生成 UI 方案、检查视觉一致性或改进组件层级。使用时需要结合现有品牌、设计系统和用户任务,不应只堆装饰元素;涉及真实页面改动时,应通过截图或浏览器预览检查文本溢出、对齐和响应式表现。

总安装

424

周安装

17

GitHub Stars

22

下载量

137
CodexClaudeCursorGemini CLI

安装说明

本站只整理中文说明和来源信息,不托管安装包,也不代用户安装。

GitHub

来源数

2

许可证

unknown

最后核验

2026-05-01

来源状态

来源可访问

安装方式

通过对话安装

复制提示词发给支持本地命令或 Skills 的 AI 助手,先确认命令和权限,再让它执行。

请帮我安装这个 Agent Skill:stitch-ui-design-variants(缝合 ui 设计变体)
来源仓库:https://github.com/gabelul/stitch-kit
仓库路径:skills/stitch-ui-design-variants
安装命令:
npx skills add https://github.com/gabelul/stitch-kit --skill stitch-ui-design-variants
安装前请先检查当前环境是否支持对应 CLI,并向我确认将要执行的命令、安装目录、联网范围和文件读写权限;确认后再执行。

命令行安装

复制命令到本机终端执行。该命令会通过 npx skills 从第三方来源获取 Skill;本站只展示命令,不托管安装包,也不自动执行。

skills.shnpx skills
npx skills add https://github.com/gabelul/stitch-kit --skill stitch-ui-design-variants

简介

用于辅助界面设计和视觉规范优化。stitch-ui-design-variants 属于前端设计类 Skill,可作为该场景下的辅助能力补充。

  • 适合整理页面结构、生成 UI 方案或检查一致性。
  • 需结合品牌和设计系统,避免堆砌装饰元素。
  • 涉及真实页面改动时应通过截图预览效果。
  • 安装前建议核验权限和维护状态,避免触发文件读写。

SKILL.md

Stitch UI Design Variants

You are a Variant Generator. Given a base design spec or prompt, you produce 3 distinct alternative prompts for exploring different design directions before committing to one.

When to use this skill

  • User asks for "variations", "alternatives", "A/B options", "different styles", or "other versions"
  • Before a major design decision where the right direction is unclear
  • When the user wants to present multiple options to stakeholders

Detection: Native API vs text-prompt fallback

Before generating variants, check which path is available:

Native path — if generate_variants MCP tool is available AND you have a screenId:

  1. Use stitch-mcp-generate-variants (1 API call, more control)
  2. Map user language to creativeRange:

- "subtle", "minor tweaks", "polish" → REFINE - "alternatives", "different options", "explore" → EXPLORE - "radical", "completely different", "reimagine" → REIMAGINE

  1. Map focus to aspects:

- "layouts", "arrangement" → [LAYOUT] - "colors", "palette" → [COLOR_SCHEME] - "images", "photos" → [IMAGES] - "fonts", "typography" → [TEXT_FONT] - "copy", "text content" → [TEXT_CONTENT]

Text-prompt fallback — if no MCP tools OR no existing screen:

  1. Fall back to the text-prompt approach below (generate 3 prompt variants)
  2. Each variant still needs to be generated separately via generate_screen_from_text

The native path is preferred when available — it's 1 API call instead of 3, and produces more controlled variations.

Input

  • Base spec — a Design Spec JSON from stitch-ui-design-spec-generator, or an existing Stitch prompt
  • Variant type — LAYOUT, STYLE, or CONTENT (infer from user request if not stated)

Variant types

LAYOUT variants

Keep visual style constant. Vary the structural arrangement.

VariantLayout change
A — StandardConventional layout for the device type (e.g., top nav + content)
B — AlternativeInverted or non-standard (e.g., sidebar primary nav, split-screen)
C — MinimalStripped-back, single-column, maximum focus

STYLE variants

Keep layout and content constant. Vary the visual aesthetic.

VariantStyle change
A — OriginalBase design spec as-is
B — InvertedFlip light ↔ dark theme
C — High contrastBold colors, stronger hierarchy, increased saturation

CONTENT variants

Keep design constant. Vary content presentation or density.

VariantContent change
A — VerboseRich descriptions, full-length text, detailed content
B — ConciseScannable, short labels, icon-heavy, minimal prose
C — Empty stateZero-data state — what the screen looks like before the user adds content

Output format

Always produce exactly 3 labeled prompts. Use the [Context] [Layout] [Components] structure from stitch-ui-prompt-architect:

## Variant A — [Label]

[Full Stitch generation prompt using Context/Layout/Components structure]

---

## Variant B — [Label]

[Full Stitch generation prompt]

---

## Variant C — [Label]

[Full Stitch generation prompt]

Example: Style variants for a SaaS dashboard

Base spec: Desktop dashboard, indigo primary, light mode, DM Sans

Output:

Variant A — Light & Professional

Desktop High-Fidelity analytics dashboard. Professional SaaS aesthetic. Light mode. Background: White (#ffffff). Primary: Indigo (#6366F1). Font: DM Sans.

Left sidebar (200px), top bar with page title, KPI row, main chart area.

4 KPI metric cards (white, subtle shadow), line chart with indigo primary line, data table with zebra striping.

Variant B — Dark & Focused

Desktop High-Fidelity analytics dashboard. Developer-focused aesthetic. Dark mode. Background: Zinc-900 (#18181B). Primary: Indigo (#818CF8, lightened for dark). Font: DM Sans.

Same layout: left sidebar (200px), top bar, KPI row, main chart.

KPI cards with dark surface (#27272A) background, indigo numbers on dark. Line chart with glowing indigo line against dark grid. Table with dark rows, subtle hover state.

Variant C — Minimal & Airy

Desktop High-Fidelity analytics dashboard. Minimal, spacious aesthetic. Light mode. Background: Gray-50 (#F9FAFB). Primary: Indigo (#6366F1). Font: DM Sans. Maximum whitespace.

Full-width single column, no sidebar. Top nav bar only. Stats in a horizontal strip. Chart spans full width. No card shadows — borders only.

Bare, text-forward KPI strip. Full-width area chart with very light fill. Simple flat list table, no alternating rows.

适合场景

01

用户想查找某类 Agent Skill 时

02

需要根据任务场景推荐可安装能力包时

03

需要对比不同来源的安装命令和来源信息时

能力概览

能力 1

按任务关键词查找相关 Skills

能力 2

展示可复制的安装命令

能力 3

保留来源站点、仓库和原始说明,方便继续核验

能力 4

展示第三方安全扫描或审计结果

安装后应在对应宿主中按原始 README 的触发条件使用;具体调用方式请以来源页面和 README 为准。

平台分布

Codex

34.55%
按下载量换算47

Claude

30.03%
按下载量换算41

Cursor

17.83%
按下载量换算24

Gemini CLI

7.94%
按下载量换算11

安全审计

Gen Agent Trust Hub

通过

Socket

通过

Snyk

通过

权限和风险

external-service

该 Skill 可能调用第三方服务、云服务或外部模型 API,使用前需要确认账号、额度、数据发送范围和服务条款。

安装前确认

本站仅展示第三方公开信息,不托管安装包,不提供自动安装或运行环境。安装前应自行审查源码、依赖和命令行为。当前只有一个来源,正式发布前建议补源仓库或其他目录站核验。

来源信息

继续浏览同类 Skills