Spark Playwright page check

Verify the requested browser-visible behavior with the repository's Playwright installation. HTTP success and empty error logs are supporting evidence, not proof that client plugins or the requested interaction work.

Procedure

1. Resolve the target URL and acceptance criteria from the task and running service. Check the owning package scripts and browser installation; use the existing setup command only when the environment needs it.
2. Run a temporary probe from a package that resolves Playwright, or extend an existing browser test when the behavior warrants a regression test. Capture console errors, page exceptions, failed requests, and HTTP error responses. A completed 4xx/5xx response is not a requestfailed event.
3. Navigate and wait for an explicit application-ready element or state. Exercise the requested interaction and assert its visible result; avoid treating networkidle or a fixed delay as readiness for a long-lived event connection.
4. Capture the relevant body evidence and screenshot. View the screenshot before making a visual claim. Bound observation time and report a timeout as missing evidence rather than success.
5. Close the browser even if the probe fails and remove only temporary files created by this check. Keep them out of commits.

Evidence and authority

Report relevant errors, status, assertions, and rendered evidence, redacting credentials and private data rather than dumping logs verbatim. Distinguish page load, plugin readiness, and interaction acceptance. An empty error array only means no error was observed in the checked interval.

Browser verification does not change service trust: direct LAN access accepts only loopback or the host's discovered local interface IP literals, and remote peers still require daemon-user authentication.

Return pageStatus, consoleErrors, pageErrors, failedRequests, bodyEvidence, verdict, and blockingReasons.