By stratified ZF -Reg., it is meant the theory ZF - Reg. with restricting Separation and Replacement to stratified instances of them.
Is Stratified ZF consistent with having an internal non-trivial automorphism $j$ from $V$ to $V$?
That is we add a new primitive one place function symbol $j$ to the language of ZF, and add the axioms of $j$ being an automorphism over the whole universe of discourse, that is:
Bijectivity: $\forall x \exists y : j(x)=y \\\forall y \exists x: j(x)=y$
Isomorphism: if $\phi(x_1,..,x_n)$ is a formula with $x_1,..,x_n$ being all of its free variables, then $$\forall x_1,..,\forall x_n \ [ \phi(x_1,..,x_n) \iff \phi(j(x_1),..,j(x_n))]$$
Critical point: $\exists x: x < j(x)$
Where $``j"$ is allowed to appear in instances of the stratified separation and replacement.
If this is allowed what would be the consistency strength of such addition?