Two clicks · about 30 seconds

Watch a disclosure break.Watch the gate catch it.

A cookie overlay covers a declared AI notice in the delivered page. Replay the checked-in v0.3.0 evidence, inspect the failure, then compare the fixed fixture under the same declared checks.

No legal verdict is simulated. Every result below is limited to the configured technical observation.

  1. 1Broken page
  2. 2Evidence
  3. 3Fixed gate
art50-ci@0.3.0report summary

$ npx art50-ci audit --config examples/browser-obstruction-demo/broken.yml

One declared disclosure. One delivered page.

The label exists, but a cookie overlay covers its sampled point.

Configured technical assertions only · not a legal conclusion

Step 1 of 2

Replay the recorded broken-fixture result.

This interaction replays checked-in v0.3.0 report data and links the underlying files; it does not manufacture a browser-only result.

What this proves

The regression is observable, reproducible, and reviewable.

Same declarationThe selector, expected copy, viewport, obstruction, and accessible-name expectations do not change between runs.

Real evidenceThe checked-in v0.3.0 evidence includes portable JSON, a rendered HTML report, a screenshot, timestamps, and SHA-256 hashes for both outcomes.

Narrow boundary`PASS` means only that the configured technical condition was observed at that recorded time—not compliance or certification.

Choose the next step

Install it yourself

Free, MIT-licensed Action and CLI.

Install from Marketplace Inspect the source

Monitored implementation

€500 one-time · one product · 30 days.

We implement one GitHub Actions workflow with up to five declared checks, establish a reproducible baseline, and triage up to three distinct failed-run causes during monitoring.

Check monitored launch fit