Added Oct 28, 2026
Web Accessibility (magnus919)
Has an AI agent build accessible web interfaces the right way round: define users, assistive tech and acceptance criteria first, then implement, then verify in real browsers and screen readers.
- Ecosystem
- Agent Skills
- Skill category
- Coding
- Creator / organization
- magnus919
What it is
This is one skill from the agent-skills repo that magnus919 maintains on GitHub, a catalogue of 180 skills written in the open Agent Skills format; this entry covers web-accessibility only. The skill opens by saying accessibility is "successful task completion and recovery for disabled people, not a tool score." It gives the agent seven steps: define the task, users, supported browsers and assistive technology, input methods and failure recovery; choose native HTML first; specify every interactive state's role, name, keyboard commands and focus entry and exit; load references on semantics, keyboard focus, form errors, dialogs, composite widgets and visual and motion adaptation as needed; then verify in real browser and assistive-technology combinations and record the evidence. What sets it apart is a flat rule against issuing a WCAG conformance verdict from an automated scan, with the scope, tool versions and uncovered items of manual testing written down instead. It cites WCAG 2.2 and WAI-ARIA 1.2.
Who it's for:Frontend developers and designers who need to meet accessibility requirements. Not useful if you don't code.
Before you start: Install an agent that supports the Agent Skills format, then add this one skill with npx skills add or install the whole repo. The skill has no runtime dependency. Docs are in English. The full repo holds 180 skills on everything from actuarial modelling to game engines, so scan the list before a whole-repo install and take only what you need. The code is free under the MIT license.
Based on the project's public documentation. Not tested by AI Inspiration World.