I brought this up on the mailing list. It seems like the internals of the workbench assume that it's managing Sesame HTTP compatible databases. This makes it quite difficult for 3rd party vendors to plug into the workbench, even if they have all of the relevant Repository, RepositoryFactory, etc, implementations.
There are several places in the code, Dale noted a few in the aforementioned thread, where it's hard coded to create/use an HTTPRepository. In those instances, it would be nice if a RepositoryFactory or similar mechanism were used to make the workbench more pluggable.