Roadmap and FAQ

Roadmap (not promises)

NextWhy
Independent labelled corpus and third-party reviewReplace self-authored accuracy with outside evidence
Directory and sitemap batch scanReal projects have many pages
Consented task runner on top of the browser adapterReproducible keyboard journeys
MCP server (read-only, root allowlist)Let agents call the scanner safely
Signed attestationsMove from edit detection to authenticity
Accessible dashboardReview findings and observations together

FAQ

Does it make my site WCAG compliant? No. It finds some candidates and organises human testing.

Why does it flag placeholder-only inputs when axe-core does not? Placeholders vanish on input and are weakly supported as names. It is a deliberate rule; confirm with a human check.

Is it safe to run on a production site? It never scans remote sites. The browser adapter only opens file: and loopback URLs and never submits forms.

Does it send my HTML anywhere? No. There is no network code in the core.

Why is the benchmark 100%? Because the corpus is small and written to the tool's own specification. It is a regression guard, not an accuracy claim.

Can an AI agent mark a manual check as passed? No. The software requires a named reviewer, environment and note, and states the observation is not independently validated.

How do I cite it? See CITATION.cff.