Overview

Accessible language is easy to talk about and hard to catch. Reviewing learning copy for plain language, idioms, ableism, and inclusive framing by eye can be slow. Even a careful pass misses things.

Approach

I built /allyscan as an AI skill. It runs four checks on any draft: plain language, idioms, ableism, and identity or context. Each flag comes with WCAG reasoning, so the feedback follows a standard, not a hunch. The craft was in the prompt design: I wrote each check to explain why something is flagged and what to change, not just that it is wrong.

Solution

/allyscan returns a prioritized report, from must-fix to optional, with the reasoning behind every flag. It can review a 15-page draft in about five minutes, work that takes much longer by hand. It catches things like "blind spot" as ableist, or "look at the box on the right" as a barrier for screen-reader users.

Toolkit