@KennethLindhardt1
If I understand your question properly, you want to run a rule that generates new data about your site in 3D, but reverts to a 2D state, and then use that data to write to new attributes in the "leaf shape", from which you could export to a geodatabase format for downstream work?
I'm just diving into Python myself, hopefully someone with more experience in Python responds to you. But I think what you are looking for is python script based reporting. You can add instanced reporting on the "leaf shapes" rule attributes of your interest, write a python script to capture this data, run a python script based report, and then connect the resultant report back to an object/rule attribute.
Found this older thread with a lot of info on how to do this: https://community.esri.com/t5/arcgis-cityengine-questions/joined-reporting-for-features-from-differe...
There's a draft python script in this thread, which I think needs a few modifications based on the comments to meet its initial goal, and a lot more changes to adjust to your desired goal, but this can get you closer to your answer.
In any case, I hope to learn more about this too, so I hope to follow and see if you get better answer from someone.
Edit: I think the linked thread was about modifying an attribute of a single shape's data to react to a selection of aggregated shape's data, done post-model-generation. So not exactly the same...
