Since many LLMs are now multi-modal, it would be useful to have a command to take a screenshot of an area in JupyterLab.
This could end up being useful if the LLM uses commands that create visual artifacts, or plots in a notebook, and want to iterate on it, so it could just "look at it" to know the latest state.