As far as I understand, the existing special page reader methods do the same, i. e. they read HTML. They belong to the Site object rather than to the Page. But there is a terrible regex in them which I had to copy and modify. I copied from the reader of another special page, but I could not make the urlpath properly, so it is now a constant that points to huwiki. I will upload it for those who are not looking for elegance in the scripts. :-) It may be useful for others: looks for red (unexisting) categories in user subpages, and puts them in 'nowiki', while sending a message to the owner. This is useful because Special:WantedCategories is flooded with these and people don't feel like correcting the others, if they have to choose from so many redlinks.
I'm a very rough user of API's... so when I can't find what I search, I simply go beck to the old system: I read html.... so I did when I needed a list of WantedCategories.
I guess that this horrible confession will elicit some reaction, and the solution you'll are searching for.
Pywikipedia-l mailing list