I think it would be easy enough to check a dictionary for a prefix string, adding another char if it isn't; then recurse on the rest of the string. Backtrack, trying a longer prefix, if it doesn't pan out down the road.

In fact, a Perl6 pattern could easily do this.

In Perl5 regex, I would think the "code" assertion would do, but the docs are inconsistant and I've never used it since the docs scared me off with "experimental". The issues are (1) can it "fail", and (2) how to get the most-recient pattern's contents.