standardized kurtosis and standardized skewness


I have seen in StatGraphics that they report, besides the skewness and kurtosis, the
standardized skewness and standardized kurtosis to test for normality of the data. In the StatReporter, the software explains that if any of those values are outside the range [-2,2] then data does not follow a normal distribution.

does anybody know how to compute standardized skewness and standardized kurtosis? does anybody understand why the range [-2,2] is used (maybe an approximation of [-1.96,1.96])?