Contrast checks are easy to make look more certain than they really are. Real pages have gradients, overlays, images, pseudo-elements, shadows, and mixed backgrounds. Once you start pretending all of that collapses into a simple clean ratio, you are already drifting away from reality.
The contrast work in A11Y Cat is better than that because it carries confidence and limitation handling. The command surface distinguishes normal text, enhanced thresholds, non-text cases, and harder background situations where the result needs to stay in review territory rather than being upgraded into a bold false claim.
That is what I mean by honest tooling. The goal is not to return the biggest pile of failures. The goal is to return findings that deserve to be trusted and to make uncertainty visible when a human still needs to look.
I would rather a tester say this needs review for the right reason than ship a polished lie with a ratio next to it.
Documentation trail
COMMAND_REFERENCE.mdTECHNICAL_GUIDE.md
