Composer: refer to pasted images by number, shown inline in the input

[coderm4v]

What are you trying to do?
Paste several screenshots into one message and refer to each one in my sentence ("compare image 1 with image 2, ignore image 3").

How do you do it today?
Pasted images appear as a row of thumbnails above the text box, separate from the text. I cannot tell which one the agent will treat as "first", so I describe them by position or content ("the left one").

Where does Paseo get in the way?
Images and text are separate, so there is no way to point to a specific image from inside the sentence. The agent also does not know which attachment I mean.

What would a better workflow look like?
Each pasted image becomes a numbered chip (Image 1, Image 2, ...) inside the input at the cursor position, and the number is sent with the message so the agent knows which attachment it refers to.

Plugin angle: as far as I can tell, composer plugins can add pills, slash commands and text attachment sources, but cannot change how the input renders or intercept paste. If a reusable extension point for this would be preferable to a core change, that would work for me too.