Review screen reader announcements with AI
Check what people hear when a page loads results, completes a task or reports an error. Keep important status changes discoverable without interrupting ordinary interaction.
Review screen reader announcements with AI when an interface updates in place and visual feedback is easy to miss. The accessible-dynamic-updates skill examines what changed, what the user was doing and whether a status message or focus move is appropriate. It checks timing, duplication and stale announcements across real interaction states.
A search filter may replace twenty results with three. Announcing the new count can be useful. Reading every result again while the user is still typing can be disruptive. Repeat the interaction at different speeds and compare the spoken feedback with the result actually on screen.
What it does
- Check loading, success, empty and error feedback.
- Review whether focus stays in a usable place.
- Identify duplicate or overlong announcements.
- Test repeated changes for stale messages.
- Record which browser and assistive technology were used.
Install and use
- Install the skill and provide an authorised running interface with reproducible updates. Use test data and a real browser and screen reader for confirmation; source access is needed for repairs.
- Ask, “Review the announcements when I filter orders quickly. Confirm the final count and check whether focus stays in the filter.”
- Repeat the interactions after any repair and distinguish observed results from untested assumptions.
FAQ
Source
Made by smukh / a11y-agent-skills contributors. Original skill: accessible-dynamic-updates, published on GitHub under the MIT licence.
View on GitHub