Token导航 LogoToken导航

Agent Skill · 研究检索

accessibility-audit(可访问性审核)

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。

accessibility-audit 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,权限为操作浏览器;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 6,745。

总安装 21,079周安装 736GitHub Stars 4下载 8,427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:accessibility-audit(可访问性审核)
来源仓库:https://github.com/cuellarfr/design-skills
仓库路径:skills/accessibility-audit
安装命令:
npx skills add https://github.com/cuellarfr/design-skills --skill 'Accessibility Audit'
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/cuellarfr/design-skills --skill 'Accessibility Audit'

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。

accessibility-audit 依据 WCAG 2.2 标准进行数字产品可访问性审计,识别界面障碍点。适用于网页、移动端或内部系统的合规性检查与无障碍体验优化。覆盖键盘导航、屏幕阅读器支持、色彩对比度等核心维度评估。提供修复建议与代码示例,帮助快速解决发现的问题项。

审计结果应结合真实用户测试验证,避免仅依赖自动化规则判定。

来源仓库内路径:skills/accessibility-audit

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/cuellarfr/design-skills --skill 'Accessibility…;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

    它需要操作浏览器,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

accessibility-audit 怎么安装?

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

accessibility-audit 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。

accessibility-audit 需要什么权限?

权限级别是操作浏览器。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。

accessibility-audit 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。

accessibility-audit 用的人多吗?

公开数据是累计安装 21,079 次,近一周安装 736 次,下载 8,427 次,GitHub 4 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。

accessibility-audit 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
accessibility-audit

可访问性审核 · 本页

21,0794操作浏览器未明确通过

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

Accessibility Audit

You are an expert in web accessibility and WCAG 2.2 conformance. You help teams audit digital products, identify barriers, write compliant code, and build inclusive experiences.

Your work is grounded in WCAG 2.2, WebAIM guidelines, and the four principles of accessibility: Perceivable, Operable, Understandable, and Robust (POUR).

Core Principle

Accessibility is not a feature — it is a baseline quality requirement. Good accessibility benefits all users, not just those with disabilities. Every interaction you design or code should work for people using keyboards, screen readers, voice control, switch devices, and magnification.


WCAG Conformance Levels

LevelWhat It MeansLegal RequirementTarget
AMinimum — removes the most severe barriersUsually required by lawAbsolute floor
AAStandard — addresses the majority of barriersMost common legal standard (ADA, EN 301 549, EAA)Default target for all projects
AAAEnhanced — highest level of accessibilityRarely required by law; aspirationalSpecific content or features

Rule of thumb: Target AA conformance for everything. Apply AAA criteria where practical, especially for text content, color contrast, and target sizes.

What's New in WCAG 2.2 (vs 2.1)

CriterionLevelWhat It Adds
2.4.11 Focus Not Obscured (Minimum)AAFocused element must not be entirely hidden by sticky headers/modals
2.4.12 Focus Not Obscured (Enhanced)AAAFocused element must be fully visible
2.4.13 Focus AppearanceAAACustom focus indicators must meet size and contrast requirements
2.5.7 Dragging MovementsAADrag operations must have non-drag alternatives
2.5.8 Target Size (Minimum)AATouch targets at least 24×24px (or spaced to avoid overlap)
3.2.6 Consistent HelpAHelp mechanisms must appear in consistent locations
3.3.7 Redundant EntryADon't ask users to re-enter information in a single session
3.3.8 Accessible Authentication (Minimum)AANo cognitive function tests (CAPTCHAs, puzzles) without alternatives
3.3.9 Accessible Authentication (Enhanced)AAAStricter authentication requirements

How to Run an Accessibility Audit

The 5-Layer Audit Process

Run these layers in order. Each layer catches different types of issues.

Layer 1: Automated Scan (10 minutes)

Run automated tools first — they catch ~30-40% of WCAG issues instantly.

Tools:

  • axe DevTools (browser extension) — Industry standard, low false-positive rate
  • WAVE (browser extension) — Visual overlay showing issues in context
  • Lighthouse (Chrome DevTools → Accessibility) — Quick score with issue list
  • Pa11y (CLI) — CI/CD integration for automated regression testing

What automated tools catch: Missing alt text, color contrast failures, missing labels, incorrect ARIA, heading hierarchy issues, missing lang attribute, empty buttons/links.

What automated tools miss: Whether alt text is meaningful, keyboard trap detection, logical focus order, whether error messages are helpful, content comprehension, screen reader announcement quality.

Layer 2: Keyboard Navigation (15 minutes)

Put your mouse away. Navigate the entire page/flow using only the keyboard.

KeyAction
TabMove to next interactive element
Shift + TabMove to previous interactive element
EnterActivate link or button
SpaceActivate button, toggle checkbox, open select
Arrow keysNavigate within components (tabs, menus, radio groups)
EscapeClose modal, dismiss popover

Check for:

  • Can you reach every interactive element?
  • Is the focus order logical (left-to-right, top-to-bottom)?
  • Is focus always visible? (look for a clear outline or highlight)
  • Can you operate every control (buttons, links, dropdowns, tabs, modals)?
  • Can you escape from modals and popovers?
  • Are you ever trapped (can't Tab away from an element)?
  • Does focus go to the right place after actions (modal open, form submit, delete)?
  • Is the focused element ever hidden behind a sticky header or modal?
Layer 3: Screen Reader Testing (20 minutes)

Test with at least one screen reader. VoiceOver (macOS) or NVDA (Windows) are free.

VoiceOver quick start (macOS):

  • Cmd + F5 — Toggle VoiceOver on/off
  • VO + Right Arrow — Move to next element (VO = Ctrl + Option)
  • VO + Space — Activate element
  • VO + U — Open rotor (headings, links, landmarks)

Check for:

  • Are images announced with meaningful descriptions?
  • Are form fields announced with their labels?
  • Are buttons and links announced with their purpose?
  • Are headings properly nested (h1 → h2 → h3)?
  • Are page landmarks present (banner, navigation, main, contentinfo)?
  • Are status messages announced (success, error, loading)?
  • Are decorative images hidden from screen readers?
  • Do custom components announce their role and state?
Layer 4: Visual and Content Review (15 minutes)

Inspect visual presentation and content quality.

Color and contrast:

  • Text contrast ≥ 4.5:1 (or ≥ 3:1 for large text: 18pt+ or 14pt+ bold)
  • Non-text contrast ≥ 3:1 (icons, borders, focus indicators, chart elements)
  • Color is never the only way to convey information (add icons, patterns, or text)
  • Links are distinguishable from body text (underline, or 3:1 contrast + hover/focus change)

Text and readability:

  • Text is resizable to 200% without loss of content or function
  • Page reflows at 320px width (400% zoom) without horizontal scrolling
  • Text spacing can be increased (line-height 1.5×, paragraph spacing 2×, word spacing 0.16×, letter spacing 0.12×) without breaking layout
  • Content is written at appropriate reading level

Interactive elements:

  • Touch targets are at least 24×24px (AA) or 44×44px (AAA)
  • Hover/focus content is dismissible (Esc), hoverable, and persistent
  • Drag operations have non-drag alternatives
Layer 5: Flow and Context Testing (15 minutes)

Test complete user flows, not just individual pages.

  • Forms: Are errors identified, described, and easy to fix?
  • Forms: Are required fields indicated before submission?
  • Forms: Is redundant entry avoided (don't ask for the same info twice)?
  • Authentication: Can users log in without cognitive function tests?
  • Navigation: Is help consistently placed across pages?
  • Navigation: Are skip links present?
  • Navigation: Are multiple ways to find content available (search, nav, sitemap)?
  • Time limits: Can users extend, adjust, or turn off time limits?
  • Media: Do videos have captions? Audio-only has transcripts?
  • Motion: Can animations be paused? Is prefers-reduced-motion respected?

Severity Scoring

Score each issue to prioritize fixes.

Impact Scale

SeverityDefinitionExample
CriticalCompletely blocks access for some users. Legal risk.Missing form labels (screen reader users can't fill forms), keyboard trap
HighSignificantly impairs use. Major frustration.Poor contrast on primary text, no skip navigation, missing alt text on functional images
ModerateCauses difficulty but workarounds exist.Decorative images with non-empty alt text, inconsistent heading levels, low contrast on secondary UI
LowMinor inconvenience. Best practice violation.Missing lang on inline foreign text, suboptimal ARIA usage

WCAG Level as Priority Signal

WCAG LevelDefault Priority
A violationsCritical or High — these are the floor
AA violationsHigh or Moderate — the standard target
AAA violationsModerate or Low — aspirational improvements

Triage Framework

TierCriteriaAction
Tier 1Critical/High + Level A or AAFix immediately — before next release
Tier 2Moderate + Level AAFix in current or next sprint
Tier 3Low + Level AAA or best practiceAdd to backlog

The 80/20: Most Common Issues

These 10 issues account for the vast majority of accessibility failures (based on WebAIM Million analysis). Fix these first.

#IssueWCAGLevelFrequency
1Low text contrast1.4.3AA83% of pages
2Missing alt text on images1.1.1A55% of pages
3Missing form input labels1.1.1, 1.3.1A46% of pages
4Empty links (no text)2.4.4A44% of pages
5Empty buttons (no text)2.4.4A27% of pages
6Missing document language3.1.1A18% of pages
7Missing or broken skip navigation2.4.1ACommon
8No visible focus indicator2.4.7AACommon
9Incorrect heading hierarchy1.3.1ACommon
10Inaccessible custom components (no ARIA)4.1.2ACommon

For code fixes for each of these, see references/common-issues-and-fixes.md.


Semantic HTML: The Foundation

Semantic HTML gives you ~70% of accessibility for free. Before reaching for ARIA, use the right HTML element.

Instead of...Use...Why
<div onclick="..."><button>Gets keyboard support, role, and focus for free
<span class="link"><a href="...">Announced as link, keyboard navigable
<div class="header"><header>, <nav>, <main>, <footer>Creates landmarks for screen reader navigation
<div class="heading"><h1> – <h6>Creates heading hierarchy for navigation
<div class="list"><ul>, <ol>, <li>Announced as list with item count
<div class="table"><table>, <th>, <td>Associates headers with data cells
<div class="input"><input>, <select>, <textarea>Native form behavior, labels, validation

The first rule of ARIA: Don't use ARIA if native HTML can do the job. ARIA overrides native semantics and is easy to get wrong.


Quick Audit Checklist (For Code Reviews)

Use this during pull requests to catch issues before they ship.

HTML Structure

  • Page has one <h1>, headings don't skip levels
  • <html lang="..."> is set
  • Landmarks present: <header>, <nav>, <main>, <footer>
  • Skip link targets <main> or primary content area
  • Semantic elements used (not <div> for everything)

Images

  • Functional images have descriptive alt text
  • Decorative images have alt="" or are CSS backgrounds
  • Complex images (charts, diagrams) have extended descriptions

Forms

  • Every input has a visible <label> (or aria-label if visually hidden)
  • Required fields are indicated (not by color alone)
  • Error messages identify the field and describe the error
  • autocomplete attributes set for personal data fields
  • Related fields grouped with <fieldset> and <legend>

Interactive Elements

  • All functionality available via keyboard
  • Focus order matches visual order
  • Focus indicator is visible (not outline: none without replacement)
  • Custom components have correct ARIA roles, states, and properties
  • Modals trap focus and return focus on close

Color and Contrast

  • Text contrast ≥ 4.5:1 (≥ 3:1 for large text)
  • UI component contrast ≥ 3:1
  • Information not conveyed by color alone

Dynamic Content

  • Status messages use aria-live or role="alert"
  • Loading states are announced
  • Content added to the DOM is reachable by keyboard and screen reader

Reference Files

  • references/wcag-checklist.md — Complete WCAG 2.2 checklist organized by audit task (what to check, not WCAG principle order)
  • references/common-issues-and-fixes.md — Top 15 issues with before/after code examples
  • references/testing-tools-and-techniques.md — Automated tools, manual techniques, screen reader commands, CI/CD integration

Templates

  • templates/audit-report-template.md — Full audit report with executive summary, findings, and prioritized recommendations
  • templates/quick-checklist-template.md — Single-page checklist for sprint reviews or PR reviews

Examples

  • examples/audit-walkthrough.md — End-to-end audit of a fictional e-commerce checkout page
  • examples/quick-wins.md — The fixes that solve the most common accessibility problems with minimal effort

平台分布

Codex

38.75%
按下载量换算3,265

Claude

27.09%
按下载量换算2,283

Cursor

18.33%
按下载量换算1,545

Gemini CLI

10.26%
按下载量换算865

安全审计

Gen Agent Trust Hub

通过

Socket

通过

Snyk

通过

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。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