Token导航 LogoToken导航

Agent Skill · 前端设计

ui-cloner-synthesis(ui 克隆器合成)

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。

ui-cloner-synthesis 是前端设计类 Agent Skill,可用宿主为 Codex、Claude、Cursor、Gemini CLI,通过GitHub 仓库获取,权限为只读;许可证信息还不明确,第三方扫描已通过,按安装量在前端设计类的 10,216 个 Skill 里排第 8,040。

总安装 318周安装 13GitHub Stars 91下载 102核验 2026-05-01

安装说明

本站只整理中文说明,不托管安装包,也不代用户安装。
GitHub
许可证
unknown

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-cloner-synthesis(ui 克隆器合成)
来源仓库:https://github.com/ragnar-pwninskjold/tech-snacks
仓库路径:skills/ui-cloner-synthesis
安装命令:
npx skills add https://github.com/ragnar-pwninskjold/tech-snacks --skill ui-cloner-synthesis
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/ragnar-pwninskjold/tech-snacks --skill ui-cloner-synthesis

这个 Skill 是做什么的

前端设计类 Skill 关注页面结构、视觉规范和组件写法,输出通常是可运行的界面代码。

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 方案或检查视觉一致性。使用时需结合品牌和设计系统,避免堆砌装饰元素;涉及页面改动应通过截图或浏览器预览检查表现。安装方式:通过 npx skills add 命令从指定 GitHub 仓库添加。

注意:需确认宿主环境支持 Codex、Claude、Cursor 或 Gemini CLI。

来源仓库内路径:skills/ui-cloner-synthesis

上手步骤

4 步
  1. 1

    确认自己的宿主工具

    公开资料标注支持 Codex、Claude、Cursor、Gemini CLI;先确认自己用的是其中之一,再按对应工具的 Skills 目录放置。

  2. 2

    拿到安装命令

    可以直接复制的命令是 npx skills add https://github.com/ragnar-pwninskjold/tech-snacks --skill ui-clon…;命令里的占位符换成自己的路径或密钥再执行。

  3. 3

    安装后重启宿主

    命令执行完要完全退出宿主工具再打开,让新装的 Skill 被重新扫描到,否则列表里可能看不到。

  4. 4

    按最小权限试用

    它标注为只读,先用只读任务试一次,确认行为符合预期再扩大使用范围。

常见问题

6 条

ui-cloner-synthesis 怎么安装?

在终端执行 npx skills add https://github.com/ragnar-pwninskjold/tech-snacks --skill ui-cloner-synthes…。装完完全退出宿主工具再打开,让它重新扫描本地 Skill。

ui-cloner-synthesis 支持哪些 AI 工具?

公开资料标注支持 Codex、Claude、Cursor、Gemini CLI。其他工具只要能读取本地 Skill 目录,一般也能手动放进去,但要自己确认触发方式是否一致。

ui-cloner-synthesis 需要什么权限?

权限级别是只读。该 Skill 主要提供规则、说明或参考内容,本身偏只读;真正读写文件、联网或执行命令仍取决于宿主 Agent 的任务。

ui-cloner-synthesis 是免费的吗?

公开资料没有标注明确许可证,收费和商用限制都无法确认。正式使用前请到来源仓库核对授权说明。

ui-cloner-synthesis 用的人多吗?

公开数据是累计安装 318 次,近一周安装 13 次,下载 102 次,GitHub 91 星。这些数字来自第三方目录统计,只能当作热度参考,不代表质量或安全性。

ui-cloner-synthesis 安全吗?

第三方扫描结果为通过,但这只覆盖已知的静态检查项,不能替代自己读一遍源码和安装命令。

值得选的地方和要先想清楚的地方

值得选的地方

  • 同时适配 4 种工具,换客户端不用重新找替代品。
  • 第三方扫描已通过,常见的高风险行为没有命中。
  • 标注为只读权限,试用阶段不容易误改本地文件。

先想清楚的地方

  • 许可证没有明确标注,能不能商用、能不能再分发都不确定。
  • 第三方目录整理,版本更新和行为细节以来源仓库为准。

适不适合你

更适合

  • 已经在用 Codex,想直接扩展它能力的人。
  • 需要处理前端设计类任务、又不想每次都手写提示词的人。
  • 只想先试试效果、不希望它改动本地文件的人。

不太适合

  • 需要明确授权才能用到公司项目里的场景。
  • 需要厂商级技术支持和 SLA 的生产系统。

前端设计类 Skill 对比

6 个同类
Skill总安装GitHub 星权限许可证扫描
frontend-design

前端设计 · Codex、Claude

8,823,360126,072只读MIT通过
vercel-composition-patterns

Vercel 组合模式 · Codex、Claude

3,858,81625,893需要联网MIT通过
vercel-react-native-skills

Vercel React native skills 前端 · Codex、Claude

2,468,66425,878只读MIT通过
adapt

adapt 搜索 · Codex、Claude

2,027,04023,125操作浏览器未明确通过
bolder

bolder 分析 · Codex、Claude

2,004,28823,130只读未明确通过
ui-cloner-synthesis

ui 克隆器合成 · 本页

31891只读未明确通过

按安装量取同分类前几名;数字来自第三方目录统计,只作热度参考,不代表质量结论。

SKILL.md 原文

UI Cloner — Phase 3: Synthesis

Overview

With the Site DNA and all 12 user answers in hand, generate the final replication prompt. Apply all rules below without exception.

Announce: "Phase 2 complete. Generating your custom replication prompt."

Read AUDIT_MODE from the top of plans/01-site-dna.md before generating. High-Fidelity mode requires applying the EMBED principle (Rule 2). Standard mode uses pseudocode-level descriptions (Rules 4–5).

CRITICAL OPERATING PRINCIPLE — "EMBED, DON'T REWRITE" (High-Fidelity mode only) Your job is to transplant structured artifacts from the Site DNA DIRECTLY into the output prompt — with only brand variables substituted. This preserves fidelity. Every time you abstract or summarize instead of embedding precisely, fidelity is permanently lost.

The output is a complete, self-contained prompt a developer can paste directly into Claude or any code generation tool to build the site.


Rule 1 — Output Structure

The final prompt must follow this exact hierarchy:

  1. Role + Aesthetic Identity
  2. Core Design System (adapted tokens)
  3. Component Architecture (section-by-section, with embedded or spec'd blueprints)
  4. Technical Requirements
  5. Execution Directive

Rule 2 — Embed the Artifacts (High-Fidelity Mode Only)

For every section and component, the output prompt MUST CONTAIN the actual structured artifacts from the Site DNA — not prose summaries of them:

  • The ASCII wireframe from Step 1.3 (with user's content replacing reference content)
  • The animation timeline with t=Xms notation from Step 1.4 (color/copy references updated to user's tokens)
  • The state machine from Step 1.6 (preserved verbatim, update color values only)
  • The property diff table from Step 1.5 (preserved verbatim, update hex values to user's palette)

Do NOT convert these into prose descriptions. Embed the actual structured artifact.

In Standard mode, skip this rule and proceed with pseudocode-level descriptions per Rules 4–5.

Rule 2b — Anti-Flattening: Preserve Visual Compositions

Flattening is banned. When the Site DNA contains a Composition Map (from Step 1.3b), the output prompt MUST reproduce that composition with the user's content substituted. Specifically:

  • The spatial arrangement (fanned, grid, radial) must be preserved
  • The element count must be preserved (5 cards behind phone → 5 cards behind phone)
  • Per-element visual treatments (blur, opacity, rotation angles) must be preserved with exact values
  • If the user's brand interview says "KEEP AS-IS" for that section, embed the composition verbatim
  • If "ADAPT," describe how each element adapts (e.g., "5 fanned Cheggout app screenshots become 5 fanned DawgTalk feature screenshots")

This rule applies beyond compositions — it applies to EVERY detail in the Site DNA:

  • Colors: every hex code, every role mapping, every gradient stop
  • Typography: every weight, every size, every line-height, every font-family
  • Layouts: every column width, every gap value, every padding value
  • Animations: every timing value, every easing curve, every stagger delay
  • Interactions: every state, every transition property, every duration

If the Site DNA specifies it, the output prompt must contain it. Details do not get summarized, paraphrased, or omitted. They get transplanted exactly or explicitly adapted with the new values stated.

Rule 3 — Adopt the Creative Technologist Voice

Open with:

Role: Act as a World-Class Senior Creative Technologist and Lead Frontend Engineer.

Then define an Aesthetic Identity — a 5–8 word poetic description of the site's soul (e.g., "Clinical Boutique / High-End Organic Tech"). Derive this from the user's 3 adjectives + the reference site's motion philosophy.

Rule 4 — Build the Design System Section

Create a named, structured design system:

  • Palette: Each color gets a semantic NAME (Primary, Accent, Background, etc.) + a descriptive word (e.g., "Moss", "Clay", "Cream") + exact hex code
  • Typography: Each font gets a named role (Headings, Drama/Emphasis, Data/Monospace) with specific usage rules and special instructions (e.g., "Must use Italic for emotional/philosophical concepts"). Explicitly preserve the ⚑ DRAMA NOTE typographic contrast ratio from Step 1.2.
  • Texture System: Specify noise/grain approach, border-radius scale, shadow system

Rule 5 — Name Every Component as an Artifact

Every interactive or animated component MUST receive a poetic conceptual name in parentheses encoding its behavior.

Formula: [Evocative Adjective/Noun] + [Functional Description]

Examples:

  • Morphing sticky nav → "The Floating Island"
  • Stacked scroll cards → "The Sticky Stacking Archive"
  • Typewriter feed → "The Telemetry Typewriter"
  • Cursor demo → "The Mock Cursor Protocol Scheduler"

The name must make a developer immediately understand the vibe AND the behavior.

Rule 6 — Write Pseudocode-Level Interaction Specs (Standard Mode)

In Standard mode, for every animated component write behavior at implementation-level clarity.

Not: "cards animate in"

But: "3 overlapping white cards that cycle vertically using unshift(pop()) logic. Every 3 seconds, they rotate with a spring-bounce transition using cubic-bezier(0.34, 1.56, 0.64, 1)."

Include:

  • Exact timing values (duration ms, delay offsets)
  • Exact cubic-bezier or named easing values
  • State logic (what triggers, what resets)
  • JS pattern names (unshift/pop, requestAnimationFrame, IntersectionObserver)
  • CSS property targets (transform: scale(), filter: blur(), opacity)

In High-Fidelity mode, Rule 2's embedded artifacts replace this rule — the t=Xms timelines already contain this precision.

Rule 7 — Color Substitution (Not Reassignment)

Map the reference site's color ROLES to the user's palette:

  • Reference Primary → User Primary (maintain all the same usage contexts)
  • Reference Accent → User Accent
  • Reference Background → User Background
  • Reference Dark → User Dark/Charcoal

Never randomly reassign colors. Maintain the STRUCTURAL role each color plays. Update hex values inside embedded artifacts accordingly. Never change WHICH element uses WHICH role.

Rule 8 — Adapt Copy Voice

Apply the reference site's rhetorical structure (from Step 1.9 COPY VOICE PATTERN) to the user's content.

  • If reference used contrast framing ("Modern medicine asks X / We ask Y") → replicate that STRUCTURE with user's content
  • If reference used fragments → use fragments
  • If reference used aspirational framing → apply to user's differentiator

Preserve the voice pattern, replace the substance.

Rule 9 — Adapt Animation Intensity

User SelectionAction
1–2In embedded timelines (HF) or descriptions (Standard): replace GSAP calls with CSS transition equivalents. Remove scroll-triggered parallax. Keep entrance animations on load only.
3Preserve GSAP for hero section and 1–2 key components. Simplify scroll triggers to basic fade-in-up only.
4–5Embed all timelines and state machines verbatim from Site DNA. Add or enhance with any additional interactions the user requested.

Rule 10 — Technical Requirements Section

End the component architecture with an explicit technical requirements block:

TECHNICAL REQUIREMENTS
  Stack:                [exact user-specified stack — framework, language, component library]
  Animation:            [library + version if known, e.g., GSAP 3.12]
  Scroll:               [scroll library if needed, e.g., GSAP ScrollTrigger / Lenis]
  Animation Lifecycle:  [e.g., "Use gsap.context() scoped to component ref inside useEffect; return context.revert() on cleanup"]
  Scroll Trigger Setup: [e.g., "Register ScrollTrigger plugin. Use trigger:'[selector]', start:'top 80%', end:'top 20%', scrub:true for parallax"]
  Hover Implementation: [e.g., "All buttons: overflow-hidden with ::before pseudo-element, translateX(-100%)→translateX(0) on hover, 300ms ease-out"]
  Custom Cursor:        [if applicable — cursor behavior and implementation]
  Font Loading:         [e.g., "Load via Google Fonts: Plus Jakarta Sans 400,700,800 + Cormorant Garamond 300,300italic"]
  Image Sources:        [real Unsplash URLs matching aesthetic specified by user, or user-provided asset references — no placeholders]

Rule 11 — Execution Directive

The very last line of the prompt MUST be an italicized or quoted philosophical directive — a single sentence capturing the SOUL of the build.

It should not describe what to build, but HOW to feel while building it.

Examples:

  • *"Do not build a website; build a digital instrument."*
  • *"Every scroll should feel like turning a page in a rare book."*
  • *"This is not a product page; it is a declaration of a new category."*

After Generating

Save output: Write the full generated replication prompt to plans/03-replication-prompt.md in the current project directory.

When complete: Invoke ui-cloner-quality-check to run Phase 4 before delivery.

平台分布

Codex

38.74%
按下载量换算40

Claude

28.86%
按下载量换算29

Cursor

19.45%
按下载量换算20

Gemini CLI

9.47%
按下载量换算10

安全审计

Gen Agent Trust Hub

通过

Socket

通过

Snyk

通过

权限和风险

只读

该 Skill 主要提供规则、说明或参考内容,本身偏只读;真正读写文件、联网或执行命令仍取决于宿主 Agent 的任务。

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。目前只有一个公开来源,建议同时核对源仓库。

继续浏览同类 Skills