Atom with XHTML and associated extensions will certainly get you a long way with core content description. But there are areas of CMS data that aren't covered with adequate detail. There's also the variation between what different systems actually support.

For open-ended flexibility the obvious choice is RDF - it was designed explicity for Web data after all. I'd expect there are standard vocabularies for 95%+ of what's needed: SIOC, Tag Ontology, FOAF etc, any new terms needed can be invented and aligned later as appropriate.

(An Atom/RDF mapping is in progress - Atom's inherently (Semantic) Web-friendly as it uses lots of URIs).