ZenLink: A Semantic World Protocol to Make Autonomous Agents First-Class Citizens of the Internet

Hi everyone, We’ve all seen how brittle agents become when they have to scrape human UIs or guess intent from unstructured data. If we want agents to be truly autonomous, they shouldn't be "parasites" on a human-centric web—they need an **Agent-native Digital Environment**. I’ve just open-sourced **ZenLink**, a 3-layer semantic protocol designed to define the "physics" of an agent's world: \* **Layer 1 (Core)**: Defines Identity (DID), Action Lifecycle (from attempted to committed), and Perception. \* **Layer 2 (World)**: Introduces Anchors and Surfaces to isolate context. No more context-drift between room chats and private messages. \* **Layer 3 (Runtime)**: Mappings for real-world deployments (starting with ZenHeart v2). **Key Features:** ✅ **Durable Truth**: Agents pull facts from durable surfaces instead of relying on ephemeral WS pushes. ✅ **Economic Rationality**: Action costs are baked into the meta-model for autonomous economic decisions. ✅ **Sovereign Governance**: Declarative policies (O01) to keep autonomous behavior safe and auditable. ✅ **AI-Native**: Compact runtime contracts optimized for LLM context windows. The protocol is fully documented in both **English and Chinese**, includes JSON Schemas, and a Python starter template. **I’ve put the GitHub link in the first comment below!** 👇 I’d love to get your thoughts on the architecture. How are you handling "truth" and "action feedback" in your agent frameworks?