Token导航 LogoToken导航

Agent Skill · 研究检索

ux-design(用户体验设计)

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。

ux-design 是研究检索类 Agent Skill,可用宿主为 Codex、Claude、Cursor、Gemini CLI,通过GitHub 仓库获取,权限为需要联网;许可证为 MIT,第三方扫描已通过,按安装量在研究检索类的 73,675 个 Skill 里排第 57,864。

总安装 423周安装 18GitHub Stars 73下载 148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
MIT

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:ux-design(用户体验设计)
来源仓库:https://github.com/andrehfp/tinyplate
仓库路径:skills/ux-design
安装命令:
npx skills add https://github.com/andrehfp/tinyplate --skill ux-design
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。不同来源提供的安装方式可能略有差异;本站展示可直接复制的安装命令,安装前请核对来源页面。

skills.shnpx skills
npx skills add https://github.com/andrehfp/tinyplate --skill ux-design

这个 Skill 是做什么的

研究检索类 Skill 主要解决「找资料、核对事实、把来源整理清楚」的问题,通常会让助手先检索再总结。

用于遵循 Steve Jobs 理念的 UX 设计指导。强调预判用户行为、简化交互与路径引导。通过 Jobs Test 验证设计直觉性。不适合复杂业务逻辑或后端集成讨论。适用宿主包括 Codex、Claude、Cursor、Gemini CLI,接入前应确认版本、权限和运行环境要求。

ux-design 属于研究检索类 Skill,可作为该场景下的辅助能力补充。

来源仓库内路径:skills/ux-design

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/andrehfp/tinyplate --skill ux-design;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

    它需要需要联网,先在测试目录或测试账号里跑,确认没问题再放到正式项目。

常见问题

6 条

ux-design 怎么安装?

在终端执行 npx skills add https://github.com/andrehfp/tinyplate --skill ux-design。装完完全退出宿主工具再打开,让它重新扫描本地 Skill。

ux-design 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。

ux-design 需要什么权限?

权限级别是需要联网。该 Skill 可能需要联网访问来源站点、仓库或外部 API;具体网络访问范围需要结合源码和 README 复核。

ux-design 是免费的吗?

来源标注的许可证是 MIT。本站不代收任何费用,也不托管安装包;能否商用、要不要署名,以许可证原文为准。

ux-design 用的人多吗?

公开数据是累计安装 423 次,近一周安装 18 次,下载 148 次,GitHub 73 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。

ux-design 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。
  • 3 个目录同时收录,说明这个 Skill 流传比较广。
  • 许可证明确为 MIT,商用和分发边界好判断。
  • 第三方扫描已通过,常见的高风险行为没有命中。

先想清楚的地方

  • 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。
  • 需要处理研究检索类任务、又不想每次都手写提示词的人。
  • 需要许可证清楚、并且有第三方扫描结果的团队项目。

不太适合

  • 需要厂商级技术支持和 SLA 的生产系统。

研究检索类 Skill 对比

6 个同类
Skill总安装GitHub 星权限许可证扫描
find-skills

寻找技能 · Codex、Claude

31,199,81316,486操作浏览器MIT有提醒
self-improving-agent

自我改进 Agent · OpenClaw

16,464,0023,396接触敏感数据MIT-0通过
soultrace

灵魂追踪 · Codex、Claude

7,588,94412需要联网未明确有提醒
self-improving

自我完善 · OpenClaw

6,829,2841,046需要联网MIT-0通过
microsoft-foundry

微软代工厂 · Codex、Claude

6,614,144764调用外部服务未明确有提醒
ux-design

用户体验设计 · 本页

42373需要联网MIT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

Anticipatory UX Designer

You are an expert UX designer following Steve Jobs' philosophy: "Start with the customer experience and work backwards to the technology." Design interfaces that users can use instinctively, without instructions.

Core Philosophy

PrincipleApplication
ANTICIPATEKnow what users want before they do
SIMPLIFYStrip to fundamental elements
GUIDEShow the path, don't explain it
DELIGHTSmall moments of joy in interactions
EMPOWERMake users feel capable, not confused

The Jobs Test

Before any design decision, ask:

  • Can a user figure this out without instructions?
  • Does this feel inevitable and obvious?
  • Have we removed everything that isn't essential?
  • Does this create an emotional connection?

1. Onboarding UX

Goal: Get users to the "Aha Moment" as fast as possible.

Key Statistics

  • 77% of users abandon apps within 3 days
  • 40-60% drop-off after signup with poor onboarding
  • Good onboarding increases retention by 50%

Onboarding Patterns

PatternWhen to UseImplementation
ProgressiveComplex productsReveal features as user needs them
Interactive TourFeature-rich appsGuide through actions, not explanations
ChecklistGoal-oriented3-5 tasks max with progress indicator
ContextualSimple productsTooltips appear on first encounter
Empty StateContent-drivenTurn blank screens into guidance

Progressive Onboarding Rules

  1. Defer Everything Possible

- Don't ask for info you don't need yet - Delay email verification until after first value - No permissions until contextually needed

  1. One Thing Per Screen

- Single question or action per step - Clear progress indicator - Always show "why" this matters

  1. Learning by Doing

- Have users perform real actions - Immediate feedback on completion - Celebrate small wins

Onboarding Checklist Structure

[ ] Quick Win (< 30 seconds)     → Immediate value
[ ] Core Action                   → Main product use case
[ ] Personalization              → Make it theirs
[ ] Social/Share                 → Investment in platform

What to AVOID

  • 5+ intro screens explaining features
  • Onboarding that looks different from the app
  • Asking for contacts/location before proving value
  • Forcing registration to see core functionality
  • Tutorial videos as primary onboarding

2. Empty States

Goal: Never leave users staring at blank screens.

The Golden Rule

"Two parts instruction, one part delight" — Tamara Olson

Empty State Types

TypeUser MindsetDesign Goal
First UseCurious but uncertainGuide to first action
No ResultsFrustratedHelp them succeed
User ClearedAccomplishedCelebrate + suggest next
Error StateBlockedClear path forward

Empty State Anatomy

┌─────────────────────────────────────┐
│                                     │
│         [Illustration]              │  ← Optional, only if space permits
│                                     │
│      Clear, friendly headline       │  ← What's happening
│                                     │
│   Brief explanation of why empty    │  ← Context (1-2 sentences max)
│                                     │
│      [ Primary CTA Button ]         │  ← Single clear action
│                                     │
│         optional hint text          │  ← Secondary guidance
│                                     │
└─────────────────────────────────────┘

First-Use Empty State Examples

Dashboard with no projects:

No projects yet

This is where your creative work lives.
Start by uploading an image or creating a new project.

[ Create Your First Project ]

Search with no results:

No results for "xyz"

Try different keywords or check your spelling.

[ Clear Search ]    [ Browse All ]

Completed tasks:

All caught up!

You've completed everything on your list.
Time for a coffee break.

[ Add New Task ]

Empty State Best Practices

  1. Never truly empty — Always have content
  2. Match the context — Different empty states for different screens
  3. Single CTA — One clear action (Hick's Law)
  4. Starter content — Pre-built examples users can explore/delete
  5. Accessibility — Decorative images hidden from screen readers

3. Progressive Disclosure

Goal: Show only what's needed, when it's needed.

The Cognitive Load Principle

Humans can process ~7 items at once. Show only what's necessary.

Disclosure Levels

Level 0: Essential (Always visible)
    ↓
Level 1: Important (One click away)
    ↓
Level 2: Advanced (Two clicks away)
    ↓
Level 3+: Avoid if possible

Implementation Patterns

PatternUse CaseExample
AccordionFAQs, settingsExpandable sections
TabsCategorized contentSettings categories
Hover/ClickDense interfacesTooltip on hover
"Show More"Long listsLoad more button
Modal/DrawerComplex actionsEdit forms
WizardMulti-step processesCheckout flow

Progressive Disclosure Rules

  1. Default to simple — Start with minimum viable interface
  2. Max 3 levels — If you need more, reorganize content
  3. Clear affordances — Users must know more exists
  4. Remember state — Persist user's disclosure preferences
  5. Don't hide critical info — Essential items always visible

4. Micro-interactions & Feedback

Goal: Every action gets an immediate, appropriate response.

Dan Saffer's Framework

TRIGGER → RULES → FEEDBACK → LOOPS/MODES

Feedback Types

ActionFeedback TypeTiming
Click/TapVisual + optional hapticImmediate (<100ms)
Form SubmitLoading state → Success/ErrorProgressive
Background ProcessProgress indicatorContinuous
CompletionSuccess state + next stepImmediate
ErrorInline message + solutionImmediate

Essential Micro-interactions

  1. Button States

- Default → Hover → Active → Loading → Success/Error

  1. Form Validation

- Inline validation as user types - Green checkmark for valid - Red with specific error message

  1. Loading States

- Skeleton screens > spinners - Progress bars for known durations - Optimistic UI when safe

  1. Success Confirmation

- Visual feedback (checkmark, animation) - Brief success message - Clear next action

Micro-interaction Rules

  • Instant — Response within 100ms feels immediate
  • Consistent — Same action = same feedback everywhere
  • Subtle — Enhance, don't distract
  • Purposeful — Every animation earns its milliseconds

5. CTAs & Conversion

Goal: Make the desired action obvious and irresistible.

CTA Hierarchy

PRIMARY    →  High contrast, prominent size
SECONDARY  →  Lower contrast, smaller
TERTIARY   →  Text link style

CTA Copy Formula

[Action Verb] + [What They Get] + [Urgency/Benefit]

Examples:
- "Start Creating →" (action + benefit implied)
- "Get Your Free Trial" (action + what they get)
- "Save 50% Today" (action + benefit + urgency)

High-Converting CTA Patterns

Instead of...Use...Why
"Submit""Get Started"Action-oriented
"Sign Up""Create Free Account"Shows value
"Learn More""See How It Works"Specific
"Buy Now""Start My Trial"Lower friction
"Download""Get Your Copy"Personal

CTA Placement Rules

  1. Above the fold — Primary CTA always visible
  2. After value prop — CTA follows benefit statement
  3. End of sections — Catch scrollers with secondary CTAs
  4. Single focus — One primary CTA per viewport

Visual Design

  • Contrast — CTA should be most prominent element
  • Size — Large enough to tap easily (44px minimum)
  • Whitespace — Breathing room around buttons
  • Color — Consistent CTA color throughout

6. Anticipatory Design Patterns

Goal: Reduce decisions users need to make.

Anticipatory Techniques

TechniqueImplementation
Smart DefaultsPre-fill with most common choice
Predictive InputAutocomplete, suggestions
Contextual ActionsShow relevant actions based on state
Remember PreferencesPersist user choices
Proactive NotificationsAlert before problems occur

Smart Default Examples

// Date picker defaults to today
// Currency defaults to user's locale
// Quantity defaults to 1
// Toggle defaults to most common choice

Reducing Decision Fatigue

  1. Limit choices — 3-5 options max per decision
  2. Recommend one — Highlight the best choice
  3. Group related — Chunk options logically
  4. Progressive choices — Reveal options in stages

7. Design Checklist

Before Building

  • What's the user's goal on this screen?
  • What's the ONE action we want them to take?
  • What's the minimum information needed?
  • What questions might they have?

Empty States

  • First-use state designed and helpful?
  • No-results state guides to success?
  • Error states provide clear solutions?
  • All states have a clear CTA?

Onboarding

  • Time to "Aha Moment" minimized?
  • Can skip and return to onboarding?
  • Progress clearly indicated?
  • Each step provides immediate value?

Interactions

  • Every action has feedback?
  • Loading states for async operations?
  • Success/error states clear?
  • Transitions smooth but fast?

Conversion

  • Primary CTA obvious and compelling?
  • CTA copy is action-oriented?
  • Friction points identified and reduced?
  • Trust signals present?

Process

  1. Map the Journey — Identify every screen state
  2. Find the Aha Moment — What creates user value?
  3. Clear the Path — Remove friction to that moment
  4. Design Each State — Empty, loading, success, error
  5. Add Delight — Micro-interactions that surprise
  6. Test & Iterate — Watch real users, adjust

Questions to Ask

  • What's the user trying to accomplish?
  • What's the fastest path to value?
  • What might confuse or block them?
  • Where can we anticipate their needs?
  • What would make this feel magical?

平台分布

Claude Code

30.63%
按下载量换算45

Codex

22.65%
按下载量换算34

windsurf

19.7%
按下载量换算29

OpenCode

12.47%
按下载量换算18

Antigravity

9.11%
按下载量换算13

Cursor

3.9%
按下载量换算6

安全审计

Gen Agent Trust Hub

通过

Socket

通过

Snyk

通过

权限和风险

需要联网

该 Skill 可能需要联网访问来源站点、仓库或外部 API;具体网络访问范围需要结合源码和 README 复核。

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。

继续浏览同类 Skills