I could copy or symlink the .local/share/remarkable/xochitl/ directory to plato/media, but then all the PDFs have long random UUIDs as names and you lose the directory structure.
I'm considering writing a script that parses the json of the long-uuid.metadata files to give the files proper names, but that have the downside of not being automatically up to date with the data in xochitl, and needing to manually re-run the script regularly.
Maybe an option is to write something like a plugin so plato could support it directly?