By repository do you mean to say something that is part of your data access layer? If so then what I do is to hook up a clean database as part of my build process (using Nant). This way when I run my build, my clean db is hooked up, any update scripts i have are ran against it to bring it up to speed, then all my unit tests are ran against my code, then my repository tests are ran to insure that my DAL is working as expected, then my db is rebuilt (essentially reset to normal), and then I am ready to go. This way I can pump in and out as much data as I like through my repository to make sure that all of the functions work there...without impacting my day to day development db/data.
If you just run tests on the your working db then you run into the problem that the data may change which might break your tests. If as part of your tests you pump known data in, and then run tests on your repository, the outcome is assumed to be known and should not change over time. This makes your test more likely to endure through time.
Hope this is what you meant!