Autonomous Node with Vision Agent. VisionAgent hook runs (see log — "Extracting content from images") but {{turn.Visionagent.content}} comes back empty, so the node never sees the extracted text. I have "Extract from Incoming Images" on and the Extract Content from Image card added. Could a before_incoming_message hook (I have desk-hitl and an inject_learn hook running) be clearing the turn state before the node reads it? How do I capture the vision output reliably? https://cdn.discordapp.com/attachments/1540211197587099758/1540211198027374623/image.png?ex=6a8920f5&is=6a87cf75&hm=c9c8538e28510858a191556f5b2169dd882aa06df603df40ae63b550b1cff53e&