Validation results
LIMITED EVIDENCE. The published result is a local execution snapshot, not an AI-accuracy or cross-platform benchmark.
Frontguard publishes raw validation artifacts in validation/results-v0.2.md rather than unverified marketing percentages.
Published run
What it proves
- The local CLI route discovery, Playwright rendering pool, and pixel/SSIM comparison pipeline completed 39 of 43 route rechecks across the 2 fixture applications that booted cleanly on the test host.
- Repeated local rechecks on the same host with deterministic stabilization enabled produced consistent pass/fail outcomes.
What remains unvalidated
The run did not seed known visual regressions, measure model precision or recall, compare operating systems, or prove public Action and hosted workflows.
