The purpose of this app is to help the user learn what to expect on a QQ plot when the data is known to have come from a normal distribution.  Repeatedly simulating new data shows how much deviation from the diagonal line (which represents agreement between a perfect theoretical normal distribution and the distribution of some data) is to be expected from just random chance.  Adjust the skew and/or tail length (aka kurtosis) as desired, and see what happens to the QQ plot.

Notice that, for smaller sample sizes, QQ plots of data from a normal distribution are often hard to distinguish from those of data from skewed and heavy-tailed distributions.  As sample size increases, deviations from normality become more obvious.

Also notice that, even under normality, data toward the ends of the QQ plot line (representing the tails of the distribution) always fall further from the line than data in the center.   This reflects a phenomenon that is easier to see in the histogram: data generated from a normal distribution will better approximate that normal distribution's center than its tails.  Stated differently, larger sample sizes are needed to correctly "fill out" the tails of the distribution.  Seeing a handful of data points veering off at the ends of the diagonal line is not a sign that the data are non-normal.