We propose a multivariate normality test against skew normal distributions using higher-order loglikelihood derivatives which is asymptotically equivalent to the likelihood ratio but only requires estimation under the null. Numerically, it is the supremum of the univariate skewness coefficient...