Hi everyone.. Do we have a tool allowing agentic vision in agent-sdk? Gemini-3-flash is capable of selectively focusing on image elements by specifying coordinates, it seems. If there is a tool with PIL backend for image crop, it would be helpful for image analysis heavy workflows.

I believe the file read skill supports reading image files by default?

Hi Graham, I haven't given a python execution environment for my agent yet. Agent is already loaded with so many other responsibilities. Hence I didn't think about this possibility you mentioned. Let me try it out once..