We prove a multivariate central limit theorem for the numbers of critical points above a level with all possible indexes of a non-necessarily isotropic Gaussian random field. In particular, we discuss the non-degeneracy of the limit variance-covariance matrix. We extend, to the non-isotropic framework, known results by Estrade & León and Nicolaescu for the Euler characteristic of an excursion set and for the total number of critical points of Gaussian random fields. Furthermore, we deduce the almost sure convergence of the normalized (by its mean) number of critical points above a level with any given index and, in particular, of the Euler characteristic of an excursion set. Though we use the classical tools of Hermite expansions and Fourth Moment Theorem, our proof of the non-degeneracy of the limit variance-covariance matrix is completely new since we need to consider all chaotic terms.