barrierlab-agent-pluginv0.1.0

Limits

What the plugin does not do, and what a report cannot tell you.

  • One page per run. No site crawl, no baseline and no comparison between runs yet.
  • Static analysis only. What needs a browser (scripted content, focus, layout) is reported as not checked or not run, never as passed.
  • Keys are local to a report. Findings are keyed by rule id and position. This is not the engine’s cross-run FindingId, which barrierlab --json does not print.
  • Fixes are proposed, not applied. The report shows diffs.
  • No HTML conformance findings. Results of html-conform are not part of the report.
  • No score and no verdict. The report counts what the engine counted and names WCAG criteria. It does not say whether a page conforms or meets a law.
  • Claude Code only. The commands use the plugin format of Claude Code.

Planned phases are tracked in casoon/barrierlab#127.

Edit this page on GitHub · Docs for v0.1.0