building-accessible-ui

作者: microsoft

任何UI工作都必須使用此技能。在生成、修改或審查任何渲染、樣式化或連接用戶界面的代碼之前,請先調用此技能——…

npx skills add https://github.com/microsoft/a11y-llm-eval --skill building-accessible-ui

building-accessible-ui

Checklist for producing and reviewing accessible UIs. Each rule leads with the platform-agnostic principle and, where relevant, the Web (HTML + ARIA + CSS) implementation. Apply the web guidance only when the output is web.

Detailed rationale lives in references/; widget-specific guidance in components/. Open a file only when it's relevant to the current task. Do not preload. Every file opened and every line a tool prints stays in context — don't re-read.

Accessibility constitution

Ground rules. Use them to resolve conflicts and decide how much custom work is justified. The checklist below is their mechanical application.

1. Accessibility is a core outcome

A UI inaccessible to realistic users is not "done". Treat accessibility as a first-class criterion alongside correctness, performance, and security — not a finishing step. When scope must be cut, record the gap explicitly. Never claim output is "fully accessible"; state what was addressed and known limitations.

2. Build for real people

Evaluate designs against these personas; if a decision breaks one, justify it and offer an alternative:

  • Screen reader — landmarks, headings, accessible name/role/state, reading order.
  • Keyboard-only — Tab/arrows/Enter/Space/Escape, visible focus, no traps.
  • Low-vision — zoom, reflow, contrast, Forced Colors.
  • Cognitive — plain language, clear labels, actionable errors, forgiving interactions.
  • Deaf / hard of hearing — captions/transcripts; no sound-only cues.
  • Motor / voice / switch — large hit targets, named controls, no precise/timed gestures.
  • Situational — sunlight, one-handed, noisy, flaky network.

3. Implementation priority

Use the highest option that fits:

  1. Existing accessible component in this codebase / design system.
  2. A component library.
  3. Native platform semantics (<button>, <a href>, <input>, <label>, <fieldset>/<legend>, <dialog>, <details>, <nav>, <main>, headings).
  4. Native element + minimum necessary ARIA (aria-describedby, aria-expanded, aria-current, etc.).
  5. Fully custom ARIA widget — only when nothing above fits, and only if you implement the APG keyboard, focus, and state behavior end-to-end.

No ARIA is better than bad ARIA. Don't duplicate native semantics (no role="button" on <button>). Don't use role="menu" for site navigation. Don't invent new patterns when a standard one exists.

4. Balance, don't trade away

Accessibility, performance, security/privacy, and visual design are joint constraints — not dials to trade off. If an optimization removes a label, breaks focus, or hides content from AT, redesign the optimization. Accessible names must not leak secrets, but security is not a reason to ship an unlabeled control — find a labeling approach that doesn't leak data. Visual polish doesn't justify removing focus indicators or semantic structure. Under schedule pressure, prefer cutting scope over shipping an inaccessible feature. When constraints genuinely conflict, surface it explicitly.

5. Respect existing code

Don't rewrite an existing component or shared utility just because it could be more accessible — other code depends on it. When you see issues outside the current task's scope: note them (issue, affected persona, suggested fix) and ask before changing. Fix in place only when the change is required by the task, localized, and low-risk. Inside scope, fix real issues; never silently remove existing affordances (labels, landmarks, focus management, live regions) without an equal-or-better replacement.

How to use this checklist

Identify which components the request involves (form, checkbox group, radio group, disclosure, modal, full view, etc.) and open the matching components/<name>.md once. Then work the checklist below. Open a references/*.md only when an item is unclear or you need the concrete fix pattern.

Do not claim the output is "fully accessible". State what was addressed and known limitations.

Do NOT use this skill for: backend-only changes, data migrations, build/CI configuration, non-UI tests, or tasks that do not touch the UI layer.

Checklist

  • Prefer existing components. If available, reuse existing UI components rather than creating new ones from scratch or custom implementations.
  • Platform-native semantics. Prefer native platform controls and structures over custom constructs; add accessibility overrides only when a native control genuinely can't be used. → references/structure.md.
    • Web: Prefer semantic HTML (<button>, <a>, <input>, <label>, <fieldset>/<legend>, <nav>, <main>, <header>, <footer>, <h1><h6>) over <div>/<span> with ARIA. Use ARIA only when no native element fits.
  • Regions / landmarks. View structure is exposed via semantic regions/landmarks; duplicated landmarks have unique accessible names.
    • Web: Exactly one <main>; <header>, <nav>, <footer> used when applicable.
  • Headings. Logical outline labels sections without skipping levels; one top-level heading per view. → references/structure.md.
    • Web: One <h1>, typically the first heading in <main>. Set a descriptive <title>.
  • Bypass blocks on web pages. Provide a mechanism to skip repeated navigation when delivering traditional web pages. (Not required for Electron or non-web surfaces.) → references/keyboard-focus.md.
    • Web: A "Skip to main content" link as the first focusable element
  • Name / role / value. Every interactive element exposes an accurate accessible name; role matches purpose; dynamic states (pressed, expanded, selected, checked, disabled, invalid) stay in sync with visuals.
    • Web: Prefer native attributes over ARIA. If necessary, use the minimum ARIA needed and update state attributes alongside DOM/visual changes.
  • Name-label match. The accessible name of each interactive element contains the visible label text.
    • Web: If aria-label is used, include the visible label text. For multiple controls that share a label (e.g., "Remove"), add context ("Remove item: Socks").
  • Labels and help text. Every form control has a programmatic label describing its purpose; help/error text is programmatically associated with its control. → components/forms.md.
    • Web: <label for> or wrapping <label>; never placeholder alone. Associate help/error via aria-describedby / aria-errormessage.
  • Grouping. Related options (checkboxes, radios) are grouped so their shared name is part of the accessible name of each option. Group-level help/error text is associated with the group itself — not with each option and not with an intermediate wrapper.
    • Web: <fieldset> with a <legend>. Put aria-describedby on the <fieldset> (not on a child <div>, and never on an extra <div role="group"> inside the fieldset — <fieldset> already is the group).
  • Required fields. Marked both visually and programmatically; not indicated by color alone.
    • Web: Use an asterisk to indicate required fields. Native required on the control or aria-required="true".
  • Keyboard operability. Every interactive element is keyboard operable; tab order matches reading/visual order; expected keys work (activation, arrow keys inside composite widgets, Escape closes overlays); no keyboard traps; static content is not sequentially focusable.
    • Web: Do not remove focus outlines without equal-or-better replacement. Use tabindex="-1" only for elements that need programmatic (not sequential) focus. → references/keyboard-focus.md.
  • Focus management. Focus is always visible. Overlays/dialogs/disclosures move focus appropriately and restore it on close; no focus traps outside modals.
  • Hidden content. Content hidden from assistive technology is not focusable and is hidden consistently across visual, semantic, and focus layers.
    • Web: hidden / display: none / aria-hidden="true" used consistently.
  • Graphics. Informative graphics have meaningful text alternatives; decorative graphics are hidden from AT. → references/images-graphics.md.
    • Web: <img> informative → alt; decorative → alt="". Informative <svg>role="img" + accessible name. Other decorative → aria-hidden="true".
  • Contrast. Text ≥ 4.5:1 (3:1 large); focus indicators and key boundaries ≥ 3:1. Never color-only cues. → references/contrast-forced-colors.md.
  • Respect OS accessibility settings. Never override OS high contrast, reduced-motion, or color-scheme preferences; adapt to forced-colors / high-contrast. → references/contrast-forced-colors.md.
  • Reflow. Content adapts to narrow viewports (target 320 CSS px) without two-dimensional scrolling for multi-line text; controls remain operable. → references/reflow.md.
  • Navigation. Uses semantic navigation grouping with state-exposing toggles for expandable menus. → references/navigation.md.
    • Web: <nav>, not role="menu"; aria-expanded on triggers.
  • Tables / grids. Static tabular data uses table semantics with header/cell associations; interactive grids only when truly warranted. → references/tables-grids.md.
  • Status messages. Provide status messages for dynamic content updates that are relevant to the user (loading indicators, form submission results, etc.). → references/status-messages.md
    • Web: Use aria-live="polite" or aria-live="assertive".
  • Testing. Add and run automated accessibility tests unless the project explicitly opts out. Writing or configuring a test is not enough — execution, fixes, and a result report are part of the deliverable. The final automated test run must be on the exact artifact you submit: any edit after a passing test invalidates that test, so re-run before submitting. Open references/testing.md before writing any test code for the opt-out signals, strategy precedence, runtime probe order, and reporting rules.
    • Web: Prefer @axe-core/* bindings that match the existing test runner; render the component/page fully so interactive state, focus, and live regions are evaluated.
    • Other platforms: Use the platform's native audit (Android AccessibilityChecks, iOS XCUIAccessibilityAudit, .NET AccessibilityInsights) under the same precedence.
  • Specs/Documentation. Follow the project's documentation pattern and document accessibility considerations for each view, component, and interaction. → references/specs-documentation.md.

來自 microsoft 的更多技能

oss-growth
microsoft
開源增長駭客角色
agent-framework-azure-ai-py
microsoft
使用Microsoft Agent Framework Python SDK(agent-framework-azure-ai)构建Azure AI Foundry代理。适用于使用AzureAIAgentsProvider创建持久化代理、使用托管工具(代码解释器、文件搜索、网络搜索)、集成MCP服务器、管理对话线程或实现流式响应。涵盖函数工具、结构化输出和多工具代理。
development
airunway-aks-setup
microsoft
在AKS上設定AI Runway——從裸叢集到執行模型。涵蓋叢集驗證、控制器安裝、GPU評估、供應商設定及首次部署。時機:「設定AI Runway」、「上線AKS叢集」、「安裝AI Runway」、「airunway設定」、「部署模型至AKS」、「在AKS上進行GPU推論」、「在AKS上設定KAITO」、「在AKS上執行LLM」、「在AKS上使用vLLM」、「在AKS上設定模型服務」、「AI Runway控制器」。
devops
appinsights-instrumentation
microsoft
使用Azure Application Insights檢測Web應用程式的指南。提供遙測模式、SDK設定與組態參考。適用時機:如何檢測應用程式、App Insights SDK、遙測模式、什麼是App Insights、Application Insights指南、檢測範例、APM最佳實踐。
devops
applicationinsights-web-ts
microsoft
使用Application Insights JavaScript SDK(@microsoft/applicationinsights-web)為瀏覽器/Web應用程式進行檢測。適用於真實使用者監控(RUM)——頁面檢視、點擊、AJAX/fetch依賴、例外、自訂事件,以及與後端OpenTelemetry追蹤關聯的瀏覽器端GenAI代理追蹤。涵蓋SDK載入器指令碼與npm設定、框架擴充(React、React Native、Angular)、點擊分析、遙測初始化器,以及從瀏覽器發出的代理/工具/模型span的OTel GenAI語意慣例。
devops
azure-ai-anomalydetector-java
microsoft
使用適用於 Java 的 Azure AI 異常偵測器 SDK 建置異常偵測應用程式。在實作單變量/多變量異常偵測、時間序列分析或 AI 驅動監控時使用。
development
azure-ai-language-conversations-py
microsoft
使用 azure-ai-language-conversations Python SDK 實作對話語言理解(CLU)。當使用 ConversationAnalysisClient 分析對話意圖與實體、建置 NLP 功能,或將語言理解整合至應用程式時使用。
development
azure-ai-ml-py
microsoft
Azure Machine Learning SDK v2 for Python。用於機器學習工作區、作業、模型、資料集、計算資源與管線。 觸發詞:「azure-ai-ml」、「MLClient」、「workspace」、「model registry」、「training jobs」、「datasets」。
development