Token导航 LogoToken导航

Agent Skill · 研究检索

interaction-design(交互设计)

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。

interaction-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;许可证信息还不明确,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 53,927。

总安装 519周安装 21GitHub Stars 35下载 163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:interaction-design(交互设计)
来源仓库:https://github.com/petekp/claude-code-setup
仓库路径:skills/interaction-design
安装命令:
npx skills add https://github.com/petekp/claude-code-setup --skill interaction-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。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/petekp/claude-code-setup --skill interaction-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。

用于辅助界面设计、视觉规范和交互体验优化。适合整理页面结构、生成 UI 方案或改进组件层级。需要结合品牌、设计系统和用户任务使用。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。涉及真实页面改动时应通过截图检查文本溢出和对齐。

interaction-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/interaction-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/petekp/claude-code-setup --skill interaction-d…;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

interaction-design 怎么安装?

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

interaction-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。

interaction-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。

interaction-design 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。

interaction-design 用的人多吗?

公开数据是累计安装 519 次,近一周安装 21 次,下载 163 次,GitHub 35 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。

interaction-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。
  • 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

  • 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。
  • 需要处理研究检索类任务、又不想每次都手写提示词的人。

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。

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
interaction-design

交互设计 · 本页

51935需要联网未明确通过

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

Interaction Design

Design interactions that help real people accomplish real goals with minimum friction and maximum meaning.

The Fundamental Question

Before any pattern or timing value, ask: What is the user trying to accomplish, and how does this interaction help or hinder that goal?

Interaction design is not about mechanisms. It is about helping specific people achieve specific purposes while respecting their time, attention, and humanity.

Output Contracts

Interaction Specification

## Interaction Spec: [Component Name]

### User Goal
What is the user trying to accomplish? What are the stakes?

### Conceptual Model
How should users understand this interaction? What mental model should they form?

### States & Transitions
| State | Appearance | Transition | Emotional Tone |
|-------|------------|------------|----------------|
| Default | [appearance] | — | [calm/ready/inviting] |
| Hover | [changes] | 150ms ease-out | [responsive/acknowledged] |
| Active | [pressed] | 50ms | [committed/engaged] |
| Loading | [feedback] | — | [progressing/patient] |
| Success | [confirmation] | 200ms | [accomplished/closure] |
| Error | [clear indication] | — | [recoverable/guided] |

### Keyboard & Input
- Tab: [focus behavior]
- Enter/Space: [activation]
- Escape: [reversal/dismissal]
- Touch: [gesture, target size]

### Error Prevention & Recovery
- How does design prevent errors?
- What happens when errors occur?
- How does user recover?

### Accessibility
- Screen reader announcement
- Focus management
- Reduced motion behavior

Flow Analysis

## Flow Analysis: [Journey Name]

### User Goal & Stakes
What does success mean? What does failure cost?

### Dramatic Arc
- Setup: How does user enter this flow?
- Rising action: What builds toward the goal?
- Climax: The moment of commitment/completion
- Resolution: Confirmation and next steps

### Steps
1. [Step] → [Interaction] → [User feeling]

### Friction Analysis
- Necessary friction (builds engagement/prevents errors)
- Excise to eliminate (work that doesn't serve the goal)

### Error Paths
- [Error scenario]: [Prevention] → [Recovery]

Core Principles

1. Goal-Directed Design

Users are not "operators" triggering state changes. They are people with purposes.

Ask first:

  • Who is using this? (Not abstract "users"—specific archetypes)
  • What are they trying to accomplish?
  • What excise can we eliminate? (Work that doesn't serve their goal)
  • How should they feel during and after?

The perpetual intermediate: Most users are neither novices learning nor experts optimizing. They learned enough to work and won't learn more. Design for them—make the common path fast while keeping advanced options discoverable but unobtrusive.

2. Conceptual Models

Users must form accurate mental models of how the system works. The gap between the designer's model and the user's model is where confusion lives.

The system image (what users see and interact with) must accurately convey the designer's model. When users are confused, the system image has failed—not the user.

Signifiers communicate where action is possible and how to perform it. A drag handle icon is a signifier; the ability to drag is the affordance. Design signifiers that make affordances discoverable.

3. Direct Manipulation

Make the computer disappear so users concentrate on their task.

Core properties:

  • Visible objects of interest, always accessible
  • Rapid, incremental, reversible actions
  • Immediate feedback—action and result visibly connected
  • Operations in the problem domain, not command syntax

The syntactic/semantic insight: Direct manipulation works because it operates at the semantic level (the problem domain) rather than requiring translation into syntactic commands. This reduces cognitive load by eliminating the translation step.

When direct manipulation is inappropriate:

  • Precise specification (typing "24pt" may be better than dragging)
  • Batch operations across many objects
  • Abstract operations without spatial metaphor
  • Expert users who prefer keyboard efficiency

4. Reversibility Enables Exploration

Reversibility is not a feature—it is a psychological safety mechanism that transforms a system from intimidating to inviting.

  • Undo depth: Support multi-step undo, not just single-step
  • Congruent inverses: If Cmd+B bolds, Cmd+B unbolds (same action reverses)
  • State visibility: Show that undo is available
  • Graceful degradation: When true undo isn't possible, confirm before destructive actions

5. Error Prevention Over Error Handling

Errors are usually design failures, not user failures.

Slips (execution errors): User intends correctly but acts wrong

  • Prevent with constraints, defaults, confirmation for destructive actions

Mistakes (intention errors): User forms wrong goal or plan

  • Prevent with clear conceptual models, visible system state

Design forcing functions:

  • Interlocks: Require prerequisite actions
  • Lockins: Prevent stopping mid-action when dangerous
  • Lockouts: Prevent entering dangerous states

See references/cognitive-foundations.md for error psychology and prevention patterns.

6. Feedback & System Status

Every action deserves acknowledgment. Users should never wonder "Did that work?"

Response TimePerceptionDesign Response
<100msInstantDirect result, no indicator
100ms–1sBrief delaySubtle state change (cursor, opacity)
1–10sNoticeable waitSpinner or determinate progress
>10sLong operationProgress bar, estimate, allow cancel

Optimistic UI: Update immediately, reconcile errors gracefully. But be honest—don't show success before confirming it for important actions.

7. Modality: A Necessary Evil

Modes cause errors when users forget which mode they're in. But modes also isolate cognitive scope and can reduce errors when users would otherwise forget context.

Use modeless design (inspectors, inline editing) when:

  • User needs to see results while adjusting
  • Trial-and-error is expected
  • Context switching is frequent

Accept modality when:

  • Isolating scope reduces errors (complex multi-step wizards)
  • Action is destructive and confirmation prevents accidents
  • System needs exclusive attention (authentication)

If you must use modes:

  • Make current mode highly visible (not just status bar)
  • Ensure mode indicators are near the locus of attention
  • Provide quick escape (Escape key, clear exit path)

8. Meaning and Emotional Design

Interactions are not just efficient or inefficient—they carry meaning.

Visceral level: Immediate emotional response to appearance Behavioral level: Pleasure or frustration from use Reflective level: What the experience means, how it's remembered

A loading state that builds anticipation differs from one that frustrates—same mechanism, different meaning. Consider:

  • What should the user feel during this interaction?
  • What story does this interaction tell?
  • How does completing this action transform the user's situation?

See references/goals-and-context.md for dramatic structure and emotional design.

Decision Frameworks

Activation Type

Use FrequencyRecommended ActivationRationale
ConstantAlways visible (spatial)Zero activation cost
FrequentKeyboard shortcut + visible controlFast for regulars, discoverable for others
OccasionalMenu or command paletteSaves space, still findable
Rare/DangerousMenu only, possibly nestedPrevents accidents

When to Animate

Animation serves purposes—it is not decoration.

Animate when:

  • Maintaining object constancy during transitions
  • Communicating causality (this caused that)
  • Reducing change blindness
  • Conveying emotional tone appropriate to context

Don't animate when:

  • User performs action habitually (animation becomes excise)
  • Speed is critical (expert workflows)
  • User has indicated reduced motion preference
  • Animation would obscure rather than clarify

See references/animation-timing.md for timing, curves, and emotional qualities.

Friction: Eliminate or Leverage?

Not all friction is bad. Distinguish:

Excise (eliminate): Work that doesn't serve user's goal

  • Navigating to buried menus
  • Re-entering information the system knows
  • Waiting for unnecessary animations
  • Confirming non-destructive actions

Productive friction (consider keeping):

  • Confirmation before irreversible actions
  • Deliberate pacing that prevents rushed errors
  • Moments that build anticipation or mark significance
  • Resistance that indicates "this matters"

Input & Physicality

The device shapes the interaction. A finger on glass is not a stylus is not a mouse.

Touch Targets

PlatformMinimumComfortable
iOS44×44pt48pt+
Android48×48dp56dp+
Web (touch)44×44px48px+

Gesture Vocabulary

GestureUseConsiderations
TapPrimary actionClear affordance required
Long-pressSecondary/contextNeeds discoverability hint
SwipeReveal, delete, navigateAlways provide undo; don't hide primary actions
PinchZoomMaintain focus under fingers

Two-Handed Interaction

When designing for tablets or considering desktop power users:

  • Non-dominant hand: coarse positioning, context setting
  • Dominant hand: fine manipulation, action
  • Consider: Can both hands be usefully engaged?

See references/input-and-physicality.md for device-specific patterns and haptics.

Platform Considerations

Web

  • Focus management in SPAs (trap in modals, restore on close)
  • Respect prefers-reduced-motion
  • Design for touch AND mouse—avoid hover-only interactions
  • Keyboard navigation for all interactive elements

iOS

  • Respect system gestures and scroll physics
  • Support Dynamic Type (test at largest sizes)
  • Haptic feedback for significant moments
  • VoiceOver with proper traits

Android

  • Material motion principles
  • Predictive back gesture (Android 14+)
  • Edge-to-edge with proper insets
  • TalkBack support

Desktop

  • Keyboard shortcuts with discoverability
  • Multi-window state management
  • Trackpad gestures where appropriate
  • High information density options for power users

Anti-Patterns

Breaking Direct Manipulation

❌ Modal dialog with Preview button ✓ Live preview as user adjusts

❌ Must click Apply to see changes ✓ Changes visible immediately, undo available

Hidden State

❌ Current mode only in status bar ✓ Mode visible at locus of attention

❌ Unsaved changes with no indicator ✓ Clear dirty state (title, dot, changed button)

Excise Accumulation

❌ Confirmation for every action ✓ Undo support; confirm only destructive/irreversible

❌ Animation that user must wait through ✓ Animation user can interrupt or that doesn't block

Cognitive Overload

❌ All options visible at once ✓ Progressive disclosure based on task stage

❌ Error messages that blame user ✓ Clear explanation + recovery path

Reference Files

Load these as needed for detailed guidance:

Theoretical Foundations

  • Direct Manipulation (Shneiderman): Visibility, reversibility, immediate feedback. See references/direct-manipulation.md
  • Instrumental Interaction (Beaudouin-Lafon): Tools mediating user-object relationship; degrees of indirection, integration, compatibility. See references/instrumental-interaction.md
  • Goal-Directed Design (Cooper): Personas, perpetual intermediates, eliminating excise
  • Emotional Design (Norman): Visceral, behavioral, reflective levels
  • Computers as Theatre (Laurel): Dramatic structure, meaning, engagement

平台分布

Codex

36.13%
按下载量换算59

Claude

30.71%
按下载量换算50

Cursor

19.8%
按下载量换算32

Gemini CLI

10.51%
按下载量换算17

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。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