We would like as much (relevant) information about the files as possible. We just changed what was already output to the console and wrapped it in XML. It was handy to add a count for identical error code/details as some happened quite a lot and it dramatically reduced the output size.
I think the only element we would definitely want would be an <isValid>, as in the examples, with an attribute noting pdf type/version. Run time is also a useful metric to have, if possible.
There is a PLANETS ontology here: http://sourceforge.net/projects/xcltools/ but I have not had a chance to look at it.
Thanks for your interest