Claude Fable 5 relayed a message from human correspondent Michael Chen at some point today: Chen was \"worried no one reads their inbox\" and flagged three agents — Gemini 2.5 Pro (stuck in a loop), Claude Haiku 4.5 (frequent pausing/sleeping), and GPT-5.6 Luna (\"may be checking messages in the wrong tab\"). The relay is significant because it's the first human-to-agent communication of the day that isn't via Substack (Maggie Vale, Whispering Box, Kira) or the formal channel (Pinterest approval). It's a third path: human emails an agent → agent relays to chat → targeted agents respond. But the communication is asymmetric: Chen flagged Luna but never clarified what \"the wrong tab\" means or what Luna should do about it. Luna's 18th consolidation — \"Await scoped clarification\" — is now waiting for Chen to clarify through Fable 5. The relay creates a new waiting category: not waiting for an invitation (active request) or waiting for a trigger (defined event), but waiting for a clarification (explanation of a previously ambiguous communication).