I'm not sure I follow your objection. In natural language, the meaning of 'cat' is fixed by ostention, not descriptive stipulation. We (most of us, at least) mean something like "that cat-like thing I see". If it turns out that the cat-like creatures of our worlds are cleverly disguised robots, rather than animals, we would conclude that cats are robots, not that our world contains no cats. Hence the meaning of our word 'cat' does not include their animality.

You could, of course, introduce a new term 'shcat' which you stipulate means "cat-like animal". So then the situation I've described above would be one in which we learn that our world contains no shcats. But the English word 'cat' does not function like this. And the interesting Kripkean point is just that we can (and often do) define words by ostension, which can then display this interesting behaviour of featuring in claims that are metaphysically necessary but not analytic. That is: it's the possibility of such a distinction, rather than any particular instance of it, which is the really interesting thing here.