I've been looking for something related to this. Whenever I design unit test cases, I write out the common edge cases, let's say: valid input, invalid input (wrong data type), invalid input (too long), etc, then translate them into tests. I might write them in a way that the tests are the combination of various dimensions of input, one per variable. For example, try input1=['a', 'aaa', None] x input2=[-1, 0, 1, None] for each combination knowing which I expect to pass and fail.

Often the tests are similar (write one, duplicate, slightly change the logic), and I've always felt like there has to be some kind of test generation framework out there to do this, though I haven't heard of anyone actually using one.