I don’t think there’s anything inherently wrong with it—its an interesting technique I haven’t seen before, actually, and I can see that it might be useful sometimes. But in this case, I think you’re better off just duplicating the insert(:post) in the tests that need it rather than being clever with tags. If you find that the duplication gets more complex, i would probably split my test into two different files where each test in the file uses the same setup scenario. The cognitive load in connecting the tag with your setup is kind of high for what I see as a small savings.
Just my two cents, YMMV…
