I had a look at scour, but the python was way beyond my level. At this point I have a python file that will convert a single Inkscape svg in to something with inline styles and the stroke commands inherited and the px es removed from font-size commands so the gerber script will see them and another that processes the fpz file and pulls out the layer names and connector values. I'm now working on merging those two (and changing the svg processor to pull out the connections in the svg so I can warn about missing connectors or level names in the svgs). As well I managed to get some input on the subject from El-j by posting a bug report (that wasn't really a bug report) on github in
where he supplied some pointers to work already underway in Fritzing which look interesting, although I've yet to make any sense out of travis-ci except that it tests something (what it tests or how isn't clear to me 
). Along the way I'm learning github (at least sort of so far) and will eventually post the python there although I perhaps should try and post what I have so far sooner rather than later it sounds like. I'm just about to post a couple of real bug reports against parts. The convert silkscreen python script finds at least two fatal to Fritzing part errors in pcb and another 3 or 4 that aren't legal xml but don't break Fritizing types. So far schematic and breadboard look clean. In fact it just finished and all except pcb look to be clean xml wise.
Peter