My basic line of thinking is this...
For each wiki, we have a list of accounts with bot flags. Through SUL we know which ones operate on multiple wikis.
We also have listings of gadgets for each wiki, and by mapping names / comparing codebases we can see which ones are effectively "the same" (despite the lack of global gadgets, meaning that you have a lot of per-wiki deviations).
Between these sources of data we can create a very rough draft of all the difference bots and gadgets out there. With some additional tagging / annotations we can build a more useful directory of all the bots and gadgets that are out there. But this initial task is to just aggregate the different data sources together.