Review chart accessibility with AI
Test whether somebody using a keyboard or screen reader can answer the same question as a person reading a dashboard visually, including after a filter changes.
Review chart accessibility with AI by giving the agent a real question, such as which week had the most returns. The accessible-charts-and-dashboards skill asks for the source values, units and filters before choosing an alternative way to answer it. A table works for some charts, while a short explanation or a structured interaction may serve others better.
A dashboard can pass a superficial check and still show the wrong figure. If a filter changes the plotted line but leaves the alternative table on last month, two readers receive different answers. Repeating the same filtering task exposes a data or rendering problem that an accessible name alone cannot fix.
What it does
- Identify information available only when the pointer hovers over a mark.
- Compare the values and units in the chart and its nonvisual alternative.
- Test a filter, empty result and loading state in both presentations.
- Check whether a reader can reach a control and recognise its effect.
- Record screen-reader observations separately from code-based assumptions.
Install and use
- Provide the chart’s purpose, underlying figures and a working preview. Manual input and assistive technology are needed to verify the corresponding experiences.
- Ask, “Change the month and check that the returns figure for the same week agrees in the chart and its alternative.”
- Review discrepancies against the source data and repeat the failed task after a correction.
FAQ
Source
Made by smukh / a11y-agent-skills contributors. Original skill: accessible-charts-and-dashboards, published on GitHub under the MIT licence.
View on GitHub