Skills
Capability is not readiness: the ProofAgent Index scores agents on evaluation, context, compliance, and governance before deploy
This proposes a governance readiness framework, the ProofAgent Index, that separates "the demo worked" from "this can ship," scoring four dimensions — observed behavioral evaluation, the operating context, regulatory compliance, and governance capability covering authorization, monitoring, audit, and control. Validation in healthcare and finance produced three findings that generalize past regulated domains: context engineering strongly changes reliability, capability improves behavior but does not determine readiness, and governance evidence must stay visible rather than being averaged into a single score. It ships as ProofAgent Harness, open-source infrastructure for auditable agent evaluation.
↳ Follow the thread