HJS Foundation This document defines the Cognition-Oriented Emergence (COE) protocol, a lightweight, minimalist, narrow-waist cognitive interaction protocol. COE provides a unified "cognitive interaction layer" for the fragmented ecosystem of world models. Through four core primitives -- Judge (J), Delegate (D), Terminate (T), and Verify (V) -- COE enables heterogeneous world models and AI agents to reach verifiable consensus on shared world states within a traceable, auditable, and evolvable framework. Unlike JEP, which focuses on "accountability tracing" (post-hoc audit), COE focuses on "cognitive consensus" (ex-ante / in-situ collaboration). COE does not depend on any specific world model implementation; instead, it serves as a common "trust operating system" and "collaboration grammar" for them. JEP answers "who is responsible," and COE answers "what the world is" -- both share the J/D/T/V primitive gene, forming a complete "cognition‑accountability" dual-loop architecture.