The Lorentz group $O(3,1)$ has spinor representations (actually $SL(2,\mathbb C)$, that is the universal cover of $O(3,1)$), as well known. The problem is that now, in general relativity, we want to deal with generic transformations. So we are working with G$L(4)$.
Roughly speaking, the associated Lie Algebra $\mathfrak{gl}(4)$ doesn't admit spinor representations. See for instance: No Spinor The way out is to go to a local inertial frame, in which you recover the flat space and you can define the spinors.
For supergravity I can suggest to you D. Freedman and A. V. Proeyen (Supergravity, 2012)