Home / Benchmark

Accuracy benchmark

We have not published a benchmark run yet. Rather than show numbers we can't stand behind, this page stays empty until we run Shipshape against a consented corpus and publish the aggregate, anonymized result — precision/recall with false negatives shown, not hidden.

Method: passive, browser-visible checks only; aggregate and anonymized; only live (HTTP 200) apps counted. This is transparency about our detection — not a security guarantee.