Verifiable AI-build governance you run on yourself — four modules that prove themselves with one offline command.
Four real governance modules — an autonomy gate, build discipline, model fitness, and agent-diff code review — that classify, score, and flag AI-build decisions. The trust mechanism is one command: python3 verify.py runs every module's selftest on your machine and prints the real pass counts. Not advice, not a PDF — a runnable artifact whose proof is a reproducible exit code.
Stdlib-only, no install, no network. Part of the Proof family by AkuchiS. Full method and receipts in the README on GitHub.