Thanks for raising this - we've sort of seen XBRL over the horizon but haven't needed to address it yet. DataPump's got a component VorteXML which can generate XML output that conforms to a DTD, but the desktop VorteXML Designer is a little limited (can't handle PDF input etc) so in practice you might need to prep data in Monarch before feeding VorteXML with a clean fixed width text output.
But once you have fairly simple input data (no multiple footers but you can have multiple appends) then it's just a question of mapping elements. I've not had to automate a VorteXML process in the server version but I do know someone who's done this in anger for some large clients - please do email or PM me if you'd like more info on that.
Please do keep us posted - I'd be interested to hear how you get on!