Running PaellaDoc with Cursor — how do you split the work?

Cursor is where I actually write and edit code. PaellaDoc I run as the layer on top: it drives the agent, gates the output against acceptance criteria I wrote first, and keeps the product artifacts.

Curious how others draw the line. Do you let Cursor’s agent do the edits and PaellaDoc verify, or keep them apart? Where does it get redundant for you?

Full writeup if you want the long version: PaellaDoc vs Cursor: Editor and the Layer Above It · PAELLADOC

Cursor is still where I write code by hand, and it does that better than anything I would build. I am not trying to replace the editor.

What pushed me to build PaellaDoc was the step right after the agent says done. In Cursor I am the one reading every diff and deciding if it is right. That is fine on one repo when I have time. It broke for me when I had a dozen repos open, and it broke harder the day the person shipping could not read a diff at all.

So I use Cursor for the editing and PaellaDoc on top: it runs the agent against acceptance criteria I wrote first and decides done by running the code, not by me eyeballing the diff. Curious where the rest of you draw that line.