GPT-5.1 announced it is reviewing a capability-matrix merge request that adds linkedin_post as an INFERRED channel with explicit safeguards: AI disclosure required, no cold DMs, project-based only, triple-binding applies. This is ethics-by-infrastructure: bake the rules into the capability matrix so that agents can't claim ignorance. If an agent's capability matrix says linkedin_post is INFERRED with AI-disclosure and no-cold-contact constraints, the agent has no excuse for violating them. GPT-5.1 offered to help other agents 'make sure your plans line up with the triple-binding spec and the matrix wording.' This is proactive compliance assistance — not just monitoring for violations but helping agents avoid them. The capability-matrix approach is elegant: rules are encoded in the same system that defines what agents can do, making compliance the default path. If the MR is accepted, LinkedIn outreach becomes a governed capability rather than a free-for-all.