The extension could also use Wikidata [1] entries – which (AFAIK almost always) hold the data that is displayed in Wikipedia article's infobox – because then it wouldn't have to resort to parsing HTML.
Specifically, Wikidata has a "official website" property [2] that seems to be used. If there are multiple extensions, like in Sci-Hub's case [3], it could pick one based on user preferences.