According to the anti-oppression framework, you determine what anti-oppression requires by looking at what oppressed people say. To apply this framework you have to first know who the oppressed people are. The framework provides no real guidance on that. You cannot just listen to what oppressed people say about who oppressed people are: once again, to do this you would need to know first who the oppressed people are in order to know who to listen to.
So in order to apply the anti-oppression framework, you must have an antecedent theory of what makes someone oppressed. But if you have an antecedent theory of what makes someone oppressed, then why would you need to look at what oppressed people are saying in order to know what oppression is? So the paradox, it seems, is this: to apply the anti-oppression framework you need an antecedent theory of oppression, but if you have an antecedent theory of oppression, you do not need the anti-oppression framework.