AgentRC: Portable agent declarations for governance policy enforcement

[adeelahmad]

I'm building AgentRC—an open specification for declaring AI agents as portable OCI artifacts. I think it solves a real gap in ADK's governance story.

The Agentfile packages four declarations: IDENTITY (who the agent is), CAPABILITY (what it can do), SOP (standard operating procedures), and POLICY (governance constraints). When built, these become ai.agentrc.* labels on the image manifest.

Your governance hooks need a canonical contract to enforce against. AgentRC provides that contract. A platform reads the labels, evaluates policy, and decides deny-by-default—all without coupling to any specific policy engine.

I've got a working spec (v0.1.0-draft.6) and 800+ clones in four days. The spec is at agentrc.ai, code at github.com/adeelahmad/agentrc (https://agentrc.ai) .

Curious if this fits into ADK's governance roadmap.

[surajksharma07]

The IDENTITY/CAPABILITY/SOP/POLICY split as OCI manifest labels is a neat way to make governance portable across platforms and 800+ clones in four days is a great signal. Closing this as Show and Tell since it's a standalone spec/project rather than an ADK core issue or question, that way it stays discoverable without sitting in the queue. Would genuinely love to see how it evolves and whether folks end up wiring it into ADK-based agents, keep posting updates here.