Looks like Star's answer is what you're looking for. I just wanted to say that skewness is the third central moment of a distribution (histogram), so that histogram you have will also have a skewness related to how lopsided/skewed the distribution is. A symmetrical distribution (e.g. a Normally distributed population) would have a skewness of 0. If it's skewed left it would have a negative skewness and if it's skewed right it would have a positive skewness.
I've heard of the skewness of a histogram but I've never heard of a histogram of skewnesses, though it's possible. For example you could read in 400 images and compute the distance between an actual (x,y,z) coordinate on the 3-D printed (or blow molded) physical object, and the desired (x,y,z) coordinate of the CAD model, and then get a histogram of the distances and compute the skewness of the disance errors for each image (CAD file) (based on it's histogram), then you could get a distribution of those skewnesses (over all CAD files), where the bin value is the skewness and the y value is the count.
What's puzzling me is that none of your skewnesses are more than 1, or less than 0. I mean, I guess it could happen, like if your distributions were just very slightly lopsided to the right, though the huge spikes near (but not exactly at) 1 are also puzzling.
So I just wanted to make sure that you're getting your skewnesses correct and you really have a distribution of skewnesses and not some other type of distribution, like distribution of gray levels, or particle sizes, or distance deviations between physical model and CAD model, or whatever.