agency-evidence-collector

Run the Playwright capture script against the local server URL to generate responsive, dark-mode, and interaction screenshots. Then compare each screenshot against the original specification and document every mismatch with image references.

Capture before and after screenshots of each interactive element using the automated Playwright script. If the before and after images are identical, the interaction is broken; record the result with the TESTED or ERROR status from test-results.json.

Yes, the reality-check commands inspect resources/views directories and grep Blade files alongside HTML and CSS. Screenshots are taken from the running local server regardless of the templating engine used.

First implementations almost always contain defects, so a zero-issue claim signals insufficient inspection. Defaulting to 3-5 findings forces deeper visual review and prevents fantasy reporting where broken work gets approved.

Screenshots verify visual and interactive states but cannot confirm backend logic, security, or performance under load. It also requires a running local server and the Playwright capture script to be present in the project.