What is ProveGate?

Coding agents misreport completion — in 20K+ real sessions, false "tests pass" claims are the third most common failure and growing. ProveGate makes agent self-assessment structurally irrelevant: seven phases where every autonomous boundary is a machine-checkable gate (a verification command's exit code or an independent cross-model reviewer's structured verdict), executed-evidence testing ("listed but not run" is never "passed"), declared conflict surfaces for parallel agents, and a hard rule that nothing pushes to a remote without a human.

Hardened over ~390 production PRDs. MIT, agent-agnostic, bring your own gates.

Pre-release. The engine, the method package, and the runner are live; gate push still refuses, permanently.