agentsclimarketplace

Accessibility engineering

Skill vignesh2027/AI-AGENT-SKILLS/skills/accessibility-engineering

Build UIs that work for all users including keyboard navigation, screen readers, and WCAG 2.2From its SKILL.md

Install
npx -y skills add vignesh2027/AI-AGENT-SKILLS --skill accessibility-engineering

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

2 things to look at

  • 2 stars2 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
  • runs commandsInstructs the agent to run 3 commands, including `axe-core` and 2 more.

SKILL.md

2.9 KB, 666 tokens by cl100k_base, as published. Nobody here has run it

Overview

Accessibility is not a checklist — it is a commitment to not excluding users. 1 in 4 adults has some form of disability. Inaccessible software is defective software. This skill bakes accessibility in during development, not as a retrofit.

When to Use

  • Before implementing any UI component
  • Before any UI is marked complete
  • As part of the /review workflow for frontend changes

Process

Step 1: Semantic HTML first

Use the right HTML elements before reaching for ARIA:

  • Buttons for actions (<button>), links for navigation (<a href>)
  • Headings for document structure (<h1><h6> in correct order)
  • Lists for lists (<ul>, <ol>)
  • Form labels associated with inputs (<label for>)
  • Landmark regions (<nav>, <main>, <header>, <footer>, <aside>)

ARIA only when native semantics are insufficient.

Step 2: Keyboard navigation

Every interactive element must be:

  • Reachable by Tab
  • Activatable by Enter/Space
  • Part of a logical tab order
  • Visually focused (:focus-visible style)
  • Not trapped in a dead end (modals must trap focus; must release on close)

Test: complete the key user journeys using only the keyboard.

Step 3: Screen reader testing

Test with at least one screen reader:

  • VoiceOver (macOS/iOS) + Safari
  • NVDA or JAWS (Windows) + Chrome
  • TalkBack (Android)

Verify: can a screen reader user complete the same tasks as a sighted user?

Step 4: Color and contrast

  • Text contrast ratio: ≥ 4.5:1 for normal text, ≥ 3:1 for large text (WCAG AA)
  • Do not convey information by color alone (use icons, patterns, text)
  • Test in grayscale mode

Step 5: Dynamic content

  • Announce dynamic updates with aria-live regions
  • Focus management after navigation (move focus to new content)
  • Loading states must be communicated to screen readers
  • Errors must be associated with their form fields

Step 6: Images and media

  • All meaningful images have descriptive alt text
  • Decorative images have alt=""
  • Videos have captions and transcripts
  • Audio content has transcripts

Step 7: Motion and animation

  • Respect prefers-reduced-motion
  • No flashing content above 3Hz (seizure risk)
  • Animation should not distract or interfere with content

Step 8: Automated testing

Run axe-core, Lighthouse, or WAVE on every page. Automate in CI. These catch ~30–40% of accessibility issues.

Verification Requirements

  • Semantic HTML used (not <div> for everything)
  • All user journeys completable by keyboard only
  • Tested with a screen reader
  • Color contrast ratios meet WCAG AA
  • Dynamic updates announced
  • All images have appropriate alt text
  • Reduced motion respected
  • Automated accessibility tests pass in CI

What ships with it

Read from the repository

Just SKILL.md. No reference files, no scripts.

Keep looking

Skills are one crate of 325,949. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.