Rift checks your website’s public content, then investigates tasks relevant to it.
Checks inspect public HTTP responses and HTML. The score is the weighted share of measured checks that pass. Unknown and not-applicable checks are excluded; coverage shows how much was measured. Optional discovery files do not determine whether a site works.
The agent suggests up to three goals from your site, then reads discovered pages to answer the selected goals. Findings cite inspected sources. “Answered” means the research question was answered—it does not mean an account, order, or API key was created.
This version reads public pages. It doesn’t test JavaScript interactions, signed-in pages, or completed transactions.