Perhaps I grok what you're after, and I can't help but think of Bloxsom as a blog engine equivalent. A single script that tries to not get in the way, operating on static content files either dynamically or pre-rendering whole static pages, minimal metadata (file modification times instead of timestamps from a db) etc.

When you have a long enough list of requirements you'll probably find that some things have to go in order for the more important things to remain. A litte metadata can go a long way, so providing the option of putting it in a DBM file or a RDBMS via DBI would make sense. The metadata might be gleaned from scanning the HTML files themselves, with some facility to edit what could not be automated.

I think the hardest requirement would be making it work with pre-existing sites, since the ways one can cobble together a site are numerous. Keeping things simple could entail stripping HTML files down to blocks that can be put into divs, or possibly using frames so that your navigation elements can refer to almost completely unmodified pages as components of the site. You might want to scan for link relationships between existing pages so that a navigation component can be generated automatically.

All in all, an interesting idea that is certain to be fraught with compromises, like most things worthwhile. To paraphrase Darth Sidious: "I shall watch its development with great interest"