Read this as Bulletin 17B is being revised in Bulletin C -- https://acwi.gov/hydrology/Frequency/b17c/Bulletin17C-draft-for-USGS-EditorialReview-10Apr2017.pdf#page41
The subject is an interest that I unfortunately have not dedicated much time to. I have found issues in the past in blindly applying a regional skew such as based on physiographic area where rainfall or other conditions vary substantially.
Although not tried myself, I would plot the gauged watershed boundaries with their individual skews assigned, and this may help visually or with some multivariate, krieging or other approach to decide if a regional boundaries and skew makes sense, or if some other boundary conditions (geology, watershed size, shape, valley/stream type, climate-precipitation-elevation-aspect, etc.) should be considered in setting regional boundary conditions. I would try to avoid assigning regional skews to areas where skew values are not somewhat consistent. I have not reviewed the 17C benefits to see the result of the improved analysis techniques.
As suggested in 17C, dont use 17B values until reanalyzed.
Oh, I did not mention in enough detail, but if there is substantial skew variability in region, I would use GIS for plotting skews by gauged watershed (using boundaries), and using GIS as a visual and multivariate analysis tool. You can overlay geology, rainfall, soils, etc. as well as calculate various watershed factors such as size, shape, elevation, etc. to help determine what is driving the variance in skew values. If there is substiantial variation in rainfall within the region, from experience, applying the regional skew to a forecasting equation for ungauged drainages without some other rainfall correction coefficient is likely to cause error. In some cases, selecting and analyzing stream gauging stations that are more localized and similar relative to conditions may be justifiable and should be preferable over the standard physiographic area (mountains, piedmont, coastal plain, etc.) selection process.