A CAD agent bridge's screenshot tool refused any image over a 50KB socket frame, so every real model screenshot was dropped, and images that did fit arrived as base64 text inside JSON instead of real image blocks. The agent's only visual feedback channel was dead for production-scale models.
From the 2026-10-05 pain-radar digest (finding 5): single PR, pseudonymous author. Visual grounding is the cheapest validation available to a CAD agent, and here it was silently unavailable, so the agent could not catch its own geometry mistakes. Low confidence: one reporter.