QA Evidence Engine

Highlights

How to Use

1. Start with the positioning

   Open the page and read the top block, which states that the CLI pairs equivalent Live and Staging routes, captures screenshots, traces, network, SEO, accessibility and performance evidence, and then recommends a release gate. The four cards below the buttons summarise the product as four honest statuses, eight or more add-on audits, zero fabricated results and local reproducible runs.

2. Study the verdict model

   Scroll to the Verdict model section and read the four cards in order. PASS requires successful assertions and captured evidence, FAIL covers a failed assertion or a breached threshold, INCONCLUSIVE means the evidence cannot establish an outcome, and NOT TESTED means the test never ran. The cards also note that a failure shared by Live and Staging remains a FAIL and that missing staging credentials skip cart and checkout rather than faking success.

3. Review the run sequence

   The Run it locally section shows five commands in a code block: npm install, npx playwright install chromium, and three npm run qa calls for validate, test-engine and run. The paragraph underneath explains that output lands in reports/<run-id>/ as HTML, JSON, PDF, screenshots, traces and a SHA-256 manifest, that Live remains read-only, and that staging mutations require QA_STAGING_USER and QA_STAGING_PASS.

4. Read the scope caveat

   Finish with the Operating principle section, which states that the page describes a local evidence engine and does not claim the named Live or Staging environments were certified here. It adds that a passing fixture self-test proves honest verdicts on controlled pages, not that every production path is defect-free. The footer repeats the stack and the evidence-first tagline.