Token导航 LogoToken导航

Agent Skill · 研究检索

ux-heuristics(用户体验启发法)

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-heuristics 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 3,125。

总安装 48,457周安装 2,008GitHub Stars 779下载 15,992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:ux-heuristics(用户体验启发法)
来源仓库:https://github.com/wondelai/skills
仓库路径:skills/ux-heuristics
安装命令:
npx skills add https://github.com/wondelai/skills --skill ux-heuristics
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/wondelai/skills --skill ux-heuristics

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。

根据尼尔森的 10 条启发法和 Krug 的核心原则评估界面可用性,以提高清晰度和用户信心。

涵盖尼尔森的 10 个可用性启发法(系统状态的可见性、用户控制、错误预防、一致性、识别优于召回等)以及产品应用程序和复制模式 基于克鲁格的三个定律:“不要让我思考”、点击置信度高于点击次数以及扫描行为的字数减少 包括用于导航清晰度的主干测试、用于确定问题优先级的严重性评级量表 (0–4) 以及适用于任何接口的快速诊断检查表 提供暗模式分类、启发式冲突解决方案以及常见错误和修复

来源仓库内路径:skills/ux-heuristics

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/wondelai/skills --skill ux-heuristics;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-heuristics 怎么安装?

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

ux-heuristics 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-heuristics 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-heuristics 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。

ux-heuristics 用的人多吗?

公开数据是累计安装 48,457 次,近一周安装 2,008 次,下载 15,992 次,GitHub 779 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-heuristics 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
ux-heuristics

用户体验启发法 · 本页

48,457779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

UX Heuristics Framework

Practical usability principles for evaluating and improving user interfaces. Based on a fundamental truth: users don't read, they scan. They don't make optimal choices, they satisfice. They don't figure out how things work, they muddle through.

Core Principle

"Don't Make Me Think" - Every page should be self-evident. If something requires thinking, it's a usability problem.

The foundation: Users have limited patience and cognitive bandwidth. The best interfaces are invisible -- they let users accomplish goals without ever stopping to wonder "What do I click?" or "Where am I?" Every question mark that pops into a user's head adds to cognitive load and increases the chance they'll leave. Design for scanning, satisficing, and muddling through -- because that's what users actually do.

Scoring

Goal: 10/10. When reviewing or creating user interfaces, rate them 0-10 based on adherence to the principles below. A 10/10 means full alignment with all guidelines; lower scores indicate gaps to address. Always provide the current score and specific improvements needed to reach 10/10.

Krug's Three Laws of Usability

1. Don't Make Me Think

Core concept: Every question mark that pops into a user's head adds to their cognitive load and distracts from the task.

Why it works: Users are on a mission. They don't want to puzzle over labels, wonder what a link does, or decode clever marketing language. The less thinking required, the more likely they complete the task.

Key insights:

  • Clever names lose to clear names every time
  • Marketing-speak creates friction; plain language removes it
  • Unfamiliar categories and labels force users to stop and interpret
  • Links that could go anywhere create uncertainty
  • Buttons with ambiguous labels cause hesitation

Product applications:

ContextApplicationExample
Navigation labelsUse self-evident names"Get directions" not "Calculate route to destination"
CTAsUse action verbs users understand"Sign in" not "Access your account portal"
E-commerceMatch user mental models"Add to cart" not "Proceed to purchase selection"
Form labelsDescribe what's needed plainly"Email address" not "Electronic correspondence identifier"
Error statesTell users what to do next"Check your email format" not "Validation error"

Copy patterns:

  • Self-evident labels: "Sign in", "Search", "Add to cart"
  • Action-oriented buttons: verb + noun ("Create account", "Download report")
  • Avoid jargon: "Save" not "Persist", "Remove" not "Disassociate"
  • If a label needs explanation, simplify the label

Ethical boundary: Clarity should serve users, not obscure information. Never use plain language as a veneer to hide unfavorable terms.

See: references/krug-principles.md for full Krug methodology.

2. It Doesn't Matter How Many Clicks

Core concept: The myth says "users leave after 3 clicks." The reality is users don't mind clicks if each one is painless, obvious, and confidence-building.

Why it works: Cognitive effort per click matters more than click count. Three mindless, confident clicks are far better than one click that requires deliberation. Users abandon when they lose confidence, not when they run out of patience for clicking.

Key insights:

  • Each click should be painless (fast, easy)
  • Each click should be obvious (no thinking required)
  • Each click should build confidence (users know they're on the right path)
  • Three mindless clicks beat one confusing click every time
  • Users abandon when confused, not when they've clicked too many times

Product applications:

ContextApplicationExample
Information architecturePrioritize clarity over depthShallow nav with clear labels over deep nav with vague ones
Checkout flowsMake each step obviousClear step indicators with descriptive labels
SettingsOrganize into clear categories"Account > Security > Change password" (3 confident clicks)
Search resultsLet users drill down confidentlyCategory filters that narrow results progressively
OnboardingGuide with small, clear stepsWizard with one clear action per step

Copy patterns:

  • Progress indicators: "Step 2 of 4: Shipping details"
  • Breadcrumbs: "Home > Products > Shoes > Running"
  • Confirmations at each step: "Great, your email is verified. Now let's set up your profile."
  • Clear link text: "View all running shoes" not "Click here"

Ethical boundary: Don't use extra steps to bury cancellation flows or make opting out harder. Every click should move users toward their goal, not away from it.

See: references/krug-principles.md for Krug's click philosophy and scanning behavior.

3. Get Rid of Half the Words

Core concept: Get rid of half the words on each page, then get rid of half of what's left. Brevity reduces noise, makes useful content more prominent, and shows respect for the user's time.

Why it works: Users scan -- they don't read. Every unnecessary word competes with the words that matter. Removing fluff makes important content more discoverable and pages shorter.

Key insights:

  • Happy-talk ("Welcome to our website!") wastes space
  • Instructions nobody reads should be removed
  • "Please" and "Kindly" and polite fluff add noise
  • Redundant explanations dilute the message
  • Shorter pages mean less scrolling and faster scanning

Product applications:

ContextApplicationExample
Landing pagesCut welcome copy, lead with valueRemove "Welcome to..." paragraphs
Error messagesState problem and fix, nothing more"Password too short (min 8 chars)" not a paragraph
TooltipsOne sentence max"Last 4 digits of your card" not a full explanation
Empty statesAction-oriented, minimal"No results. Try a different search."
OnboardingOne instruction per screen"Choose your interests" not a wall of explanatory text

Copy patterns:

  • Before: "Please kindly note that you will need to enter your password in order to proceed to the next step."
  • After: "Enter your password to continue."
  • Before: "We've received your message and will get back to you as soon as possible."
  • After: "Message sent. We'll reply within 24 hours."

Ethical boundary: Brevity must not mean omitting critical information. Concise disclosures for pricing, terms, and data usage are a user right.

See: references/krug-principles.md for Krug's word-cutting methodology.

4. The Trunk Test

Core concept: A test for navigation clarity: if users were dropped on any random page (like being locked in a car trunk and released at a random spot), could they instantly answer six key questions?

Why it works: Good navigation gives users constant orientation. If users can't identify where they are and what their options are, they feel lost and leave.

Key insights:

  • Users must know what site they're on (brand/logo visible)
  • Users must know what page they're on (clear heading)
  • Major sections must be visible (navigation)
  • Options at this level must be clear (links/buttons)
  • Position in hierarchy must be apparent (breadcrumbs)
  • Search must be findable

Product applications:

ContextApplicationExample
Global navPersistent site ID and sectionsLogo top-left, main nav always visible
Page headersClear, descriptive page titles"Running Shoes - Men's" not just "Products"
BreadcrumbsShow hierarchy on all inner pages"Home > Products > Shoes > Running"
Mobile navMaintain orientation in hamburger menusHighlight current section, show breadcrumbs
SearchVisible search on every pageSearch box in header, not buried in footer

Copy patterns:

  • Page titles that match the link the user clicked
  • "You are here" indicators (highlighted nav items, bold breadcrumb)
  • Section headings that orient: "Your Account > Billing" not just "Settings"
  • Footer navigation for secondary discovery

Ethical boundary: Navigation should honestly represent site structure. Don't use misleading labels to funnel users into marketing pages.

See: references/krug-principles.md for the full Trunk Test methodology.

Nielsen's 10 Usability Heuristics

1. Visibility of System Status

Keep users informed about what's happening through timely feedback. Every action needs acknowledgment — progress bars for uploads, confirmations for submissions, skeleton screens for loading. Silent failures destroy trust. Copy pattern: "Saving..." → "Saved" (immediate state transitions).

2. Match Between System and Real World

Speak users' language, not system language. Use "Sign in" not "Authenticate", "Search" not "Query." Follow real-world metaphors (trash bin, shopping cart) and natural ordering (street → city → state → zip). One term per concept, everywhere.

3. User Control and Freedom

Provide clear "emergency exits." Undo beats "Are you sure?" dialogs every time — users click through confirmations without reading. Every flow needs cancel/exit, back buttons must never break, and soft delete with undo beats permanent deletion.

4. Consistency and Standards

Same words, styles, and behaviors should mean the same thing throughout. Internal consistency (your app) and external consistency (platform conventions: logo top-left, search top-right). Pick one term per concept — "Projects" everywhere, never mixing with "Workspaces."

5. Error Prevention

Prevent problems before they occur. Constrained inputs (date pickers over text fields), autocomplete, sensible defaults, and "unsaved changes" warnings. Two error types need different prevention: slips (accidental wrong action) and mistakes (wrong intention).

6. Recognition Rather Than Recall

Minimize memory load — show options, don't require memorization. Breadcrumbs, recent searches, pre-filled fields, dropdowns with decoded values (country names, not codes). Human working memory holds ~7 items; recognition is far easier than recall.

7. Flexibility and Efficiency of Use

Serve both novices and experts. Keyboard shortcuts, touch gestures, bulk actions, saved searches, and command palettes (Cmd+K) speed up power users. Progressive disclosure keeps it simple for beginners while experts access full power.

8. Aesthetic and Minimalist Design

Every element must earn its place. Signal-to-noise ratio determines usability — when everything screams for attention, nothing stands out. Show what matters now, hide what doesn't. One primary CTA per page, not five competing ones.

9. Help Users Recognize, Diagnose, and Recover from Errors

Error messages need three parts: what happened, why, and how to fix it. Plain language always ("Connection failed" not "ECONNREFUSED"), specific ("Password must be 8+ characters" not "Invalid"), never blame the user, and preserve their input.

10. Help and Documentation

Help should be searchable, task-focused ("How to..." not technical reference), and contextual (tooltips, inline hints). Types: inline help, contextual "?" icons, searchable knowledge base, guided tours, live support.

See: references/nielsen-heuristics.md for detailed examples, product applications, copy patterns, and ethical boundaries for all 10 heuristics.

Severity Rating Scale

When auditing interfaces, rate each issue:

SeverityRatingDescriptionPriority
0Not a problemDisagreement, not usability issueIgnore
1CosmeticMinor annoyance, low impactFix if time
2MinorCauses delay or frustrationSchedule fix
3MajorSignificant task failureFix soon
4CatastrophicPrevents task completionFix immediately

Rating Factors

Consider all three:

  1. Frequency: How often does it occur?
  2. Impact: How severe when it occurs?
  3. Persistence: One-time or ongoing problem?

Common Mistakes

MistakeWhy It FailsFix
Mystery meat navigationIcons without labels force guessingAdd text labels alongside icons
Too many choicesDecision paralysis slows usersReduce to 7 plus/minus 2 items
No "you are here" indicatorUsers feel lost in the hierarchyHighlight current section in nav and breadcrumbs
No inline validationSubmit, error, scroll cycle frustratesValidate on blur with specific messages
Unclear required fieldsUsers confused about what's mandatoryMark optional fields, not required (most fields should be required)
Wall of textNobody reads dense paragraphsBreak up with headings, bullets, whitespace
Jargon in labelsUsers don't speak your internal languageUser-test all labels, use plain language
No loading indicatorsUsers think the system is brokenShow spinner, progress bar, or skeleton screen
Tiny tap targetsMobile users misclick constantlyMinimum 44x44 px touch targets
Hover-only informationMobile and keyboard users miss it entirelyDon't hide critical info behind hover states
No undoUsers afraid to take any actionProvide undo for all non-destructive actions
Poor error messages"Invalid input" tells users nothingExplain what's wrong and how to fix it
Low contrast textUnreadable for many usersWCAG AA minimum (4.5:1 contrast ratio)
Inconsistent nav locationUsers can't find navigationFixed position, same location on every page
Broken back buttonFundamental browser contract violatedNever hijack or break browser history

Quick Diagnostic

Audit any interface:

QuestionIf NoAction
Can I tell what site/page this is immediately?Users are lost and disorientedAdd clear logo, page title, and breadcrumbs
Is the main action obvious?Users don't know what to doCreate visual hierarchy, single primary CTA
Is the navigation clear?Users can't find their wayApply the Trunk Test, add "you are here" indicators
Can I find the search?Users with specific goals are blockedAdd visible search box in header
Does the system show me what's happening?Users lose trust and re-clickAdd loading states, confirmations, progress indicators
Are error messages helpful?Users get stuck on errorsRewrite in plain language with specific fix
Can users undo or go back?Users are afraid to actAdd undo, cancel, and back options everywhere
Does it work without hover?Mobile and keyboard users are excludedReplace hover-only interactions with visible alternatives
Are all interactive elements labeled?Users guess at icon meaningsAdd text labels or descriptive tooltips
Does anything make me stop and think "huh?"Cognitive load is too highSimplify -- if it needs explanation, redesign it

Heuristic Conflicts

Heuristics sometimes contradict each other. When they do:

  • Simplicity vs. Flexibility: Use progressive disclosure
  • Consistency vs. Context: Consistent patterns, contextual prominence
  • Efficiency vs. Error Prevention: Prefer undo over confirmation dialogs
  • Discoverability vs. Minimalism: Primary actions visible, secondary hidden

See: references/heuristic-conflicts.md for resolution frameworks.

Dark Patterns Recognition

Dark patterns violate heuristics deliberately to manipulate users:

  • Forced continuity (hard to cancel)
  • Roach motel (easy in, hard out)
  • Confirmshaming (guilt-based options)
  • Hidden costs (surprise fees at checkout)

See: references/dark-patterns.md for complete taxonomy and ethical alternatives.

When to Use Each Method

MethodWhenTimeFindings
Heuristic evaluationBefore user testing1-2 hoursMajor violations
User testingAfter heuristic fixes2-4 hoursReal behavior
A/B testingWhen optimizingDays-weeksStatistical validation
Analytics reviewOngoing30 minPatterns and problems

Reference Files

Further Reading

This skill is based on usability principles developed by Steve Krug and Jakob Nielsen:

About the Author

Steve Krug is a usability consultant who has been helping companies make their products more intuitive since the 1990s. His book *"Don't Make Me Think"* (first published in 2000, revised 2014) is the most widely read book on web usability and is considered essential reading for anyone involved in designing interfaces. Known for his accessible, humorous writing style and his advocacy for low-cost usability testing, Krug demonstrated that usability doesn't require a lab or a large budget -- just watching a few real users try to accomplish tasks.

Jakob Nielsen, PhD is co-founder of the Nielsen Norman Group (NN/g) and is widely regarded as the "king of usability." His 10 Usability Heuristics for User Interface Design, published in 1994, remain the most-used framework for heuristic evaluation worldwide. Nielsen has been called "the guru of Web page usability" by *The New York Times* and has authored numerous influential books on usability engineering. His research-driven approach to interface design helped establish usability as a recognized discipline in software development.

平台分布

Codex

34.5%
按下载量换算5,517

Claude

32.36%
按下载量换算5,175

Cursor

18.23%
按下载量换算2,915

Gemini CLI

9.39%
按下载量换算1,502

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