Tooling

This part covers the instruments: how to audit a page for both audiences and how to run agent-vs-page experiments yourself.

Planned chapters:

• The a13i score. A Lighthouse-style audit concept that scores a page for human accessibility and agent legibility together, and refuses to report one without the other. The pairing is the point: a tool that only scored agent legibility would be the accessibility-washing instrument this project exists to oppose.
• Test harnesses. Reproducible setups for running the corpus experiments: pinned models, page variants, scoring rubrics, and raw output capture.

Tooling code lives in this repository under MIT once it exists.