Token导航 LogoToken导航TokenDH.com
研究检索操作浏览器github未标认证来源可访问许可证需确认审计异常

ui-extractor用户界面提取器

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

494

周安装

21

GitHub Stars

公开资料未说明

下载量

173
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:ui-extractor(用户界面提取器)
来源仓库:https://github.com/alpex-ai/ui-extractor
仓库路径:skills/ui-extractor
安装命令:
npx skills add https://github.com/alpex-ai/ui-extractor --skill ui-extractor
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/alpex-ai/ui-extractor --skill ui-extractor

简介

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。

  • 适合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。
  • 使用时需结合现有品牌、设计系统和用户任务,避免堆砌装饰元素;涉及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。
  • 安装方式:github,可通过 npx skills add 命令从指定仓库添加。
  • 适用于 Codex、Claude、Cursor、Gemini CLI 等宿主环境。

SKILL.md

UI Extractor

Extract implementation specs, design systems, and component catalogs from any UI source.

What It Does

Input: A URL, screen recording, or authenticated browser tab Output: Implementation spec with components, design tokens, user flows, and Figma export

Use Cases

  • Reverse engineer a competitor's UI → Get their exact colors, typography, spacing, components
  • Document an existing app → Auto-generate component inventory and design tokens
  • Match a reference design → Compare your app against a target and get prioritized changes
  • Extract from authenticated sites → Analyze internal dashboards, Figma files, or logged-in apps

Supported Inputs

Input TypeHow to UseWhat Happens
URLAnalyze stripe.comRecords the site with Playwright, extracts frames, analyzes
Local RecordingAnalyze./recording.movExtracts frames from your file, analyzes
Browser TabAnalyze my open Chrome tabRecords your authenticated tab via Chrome MCP, analyzes

Example Commands

# Analyze any website (auto-records it)
Analyze linear.app

# Mobile or dark mode
Analyze stripe.com in mobile dark mode

# Just get the design system
Extract the design system from notion.com and export to Figma

# Compare your app to a reference
Compare ./my-app.mov with stripe.com

# Analyze an authenticated page you have open
Analyze my open Chrome tab

Output

Implementation Spec

  • Screen inventory with frame references
  • Component catalog (buttons, cards, modals, etc.)
  • Design system tokens (colors, typography, spacing, shadows)
  • User flows with triggers and transitions
  • Implementation recommendations

Design System (JSON)

{
  "colors": { "primary": { "value": "#007AFF", "usage": "Primary actions" } },
  "typography": { "heading": { "size": "24px", "weight": "600" } },
  "spacing": { "sm": "8px", "md": "16px", "lg": "24px" }
}

Export Formats

  • CSS custom properties
  • Tailwind config
  • Figma Variables (JSON)
  • Direct Figma push via MCP

Agent Instructions

Workflow Overview

  1. Resolve input → URL, local file, or browser tab
  2. Capture frames → Record (if URL/tab) or extract from video
  3. Analyze frames → Identify components, colors, typography, flows
  4. Generate output → Implementation spec, design system, or comparison

Step 1: Resolve Input

URL detected (contains .com, .io, .app, www., or http):

./scripts/record-website.sh "https://example.com" --analyze --quality default

Options: --mobile, --dark-mode, --duration <secs>

Local file provided: Validate exists, check extension (.mov, .mp4, .webm)

Browser tab requested (mentions "tab", "browser", "Chrome", "open"): Use Chrome MCP to record the tab:

  1. Get tab context
  2. Start GIF recording
  3. Scroll/hover while recording
  4. Export to ./recordings/tab-recording.gif
  5. Extract frames: ./scripts/extract-frames.sh./recordings/tab-recording.gif./frames

No path provided: Run ./scripts/detect-recording.sh to find recent recordings

Step 2: Validate Duration

ffprobe -v error -show_entries format=duration -of default=noprint_wrappers=1:nokey=1 "$VIDEO_PATH"
  • > 5 min: Error, ask user to trim
  • > 1 min: Warn about token usage
  • ≤ 1 min: Proceed

Step 3: Extract Frames

# Standard extraction
./scripts/extract-frames.sh "$VIDEO_PATH" "./frames" --quality default

# High quality for short videos (<15s)
./scripts/extract-frames.sh "$VIDEO_PATH" "./frames" --quality high

# With motion analysis
./scripts/extract-frames.sh "$VIDEO_PATH" "./frames" --include-metadata

Quality levels:

  • low (0.5fps): Videos >60s
  • default (1fps): Most recordings
  • high (2fps): Short/complex videos

Step 4: Analyze Frames

Load frames and analyze using references/frame-analysis-prompt.md:

  1. Screens: Identify distinct screens/states
  2. Components: Catalog reusable UI elements
  3. Design System: Colors, typography, spacing, shadows
  4. User Flows: Screen transitions and triggers
  5. Implementation: Component hierarchy, state hints

Step 5: Generate Output

Full Analysis (default): Output markdown implementation spec with all sections.

Design System Only: Output JSON following references/design-system-schema.md.

Export to Figma:

./scripts/figma-export.sh --design-system ./output/design-system.json --format figma-tokens

Comparison Mode:

  1. Analyze reference first
  2. Analyze user's app
  3. Output gap analysis with prioritized changes

Error Handling

File access denied (macOS security):

I can't access ~/Desktop directly. Run this to copy your recording:
cp ~/Desktop/Screen\ Recording*.mov ./recording.mov
Then: "Analyze ./recording.mov"

FFmpeg missing:

Install ffmpeg: brew install ffmpeg

Video too long:

This video is X minutes. Please trim to under 5 minutes.

Trigger Phrases

IntentPhrases
URL"Analyze stripe.com", "Record linear.app", "Extract from example.com"
Browser Tab"Analyze my open tab", "Record my Chrome tab", "Extract from this authenticated page"
Local File"Analyze./recording.mov", "Extract from this video"
Design System"Extract the design system", "Get colors and typography", "Export to Figma"
Comparison"Compare my app with...", "What's different from...", "Match this reference"

Dependencies

Required:

  • ffmpeg / ffprobe - Video processing
  • node ≥18 - Website recording

Optional:

  • jq - Figma export
  • Chrome MCP - Browser tab recording
  • Figma MCP or FIGMA_ACCESS_TOKEN - Direct Figma push

Reference Documents

  • references/frame-analysis-prompt.md - Analysis prompt structure
  • references/design-system-schema.md - Design system JSON schema
  • references/comparison-guide.md - Comparison workflow
  • references/figma-integration.md - Figma export details

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

33.8%
按下载量换算58

Claude

32.65%
按下载量换算56

Cursor

17.92%
按下载量换算31

Gemini CLI

10.66%
按下载量换算18

安全审计

Gen Agent Trust Hub

未通过

Socket

通过

Snyk

可疑

权限和风险

操作浏览器

该 Skill 可能涉及浏览器控制能力,使用时可能读取或操作网页内容,需要在受控环境中确认权限边界。

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。来源安全扫描存在 warning/failed 结果,不能写成本站确认安全。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