CSTML is just a serialization language so it doesn't have a repo. Like HTML its purpose is to embed invisible structure into a steam of visible text. BABLR is the main way we do this: producing the CSTML documents with a parser.

To the extent the law can be parsed as English you could generate CSTML documents by parsing the text of laws (or even writing laws in CSTML) but I think the "secret weapon" scenario is more likely in the present -- that a firm would use CSTML to show them effectively encode their current understanding of the structure of the text of the written law.