oabyafK would not allow moves parallel to the edge fecause the f forces the test to be opposit to the move.
Thing is, when I try it in the Sandbox, it does allow moves parallel to the edge (which you've already noted yourself).
[oW-bW-aQ] is translated into oabyaW, which also does allow moves parallel to the edge, as well as not allowing diagonal moves.
vvssoabyaK has the same result as the first one.
But I still think uing morphing is the natural solution. You want a piece that has different moves depending on where it stands, and this was what the morphing was invented for.
I can accept that as a line of reasoning, though it'd be much more "natural" if there was a way to enter it other than opening the raw code and typing it in by hand (which is sadly open to typos and other errors).
The oflabaoslabaofly sequence I suggested was supposed to test forward left, forward right, and backward of the desired move. It obviously needs to be adjusted; oabaflabalabayflK seems closer to the mark: one step opposite the desired direction to find the edge, return, turn left 45°, return, turn left 90° (returning from front left, this turns it to front right), return, turn left 45° (returning from front right, this turns it to the desired direction), toggle range from step to slide, and go. This uses the edgefinder at the back of the move, but also checks the sides to make sure that there's no edge there as well. (It checks front and diagonally because directly to the side you'll find edges if you're trying to move diagonally from the corner.) Trying it on the Sandbox, it doesn't work, so I suspect that I missed something, but hopefully my description can help you find where I went wrong there.
Thing is, when I try it in the Sandbox, it does allow moves parallel to the edge (which you've already noted yourself).
[oW-bW-aQ] is translated into oabyaW, which also does allow moves parallel to the edge, as well as not allowing diagonal moves.
vvssoabyaK has the same result as the first one.
I can accept that as a line of reasoning, though it'd be much more "natural" if there was a way to enter it other than opening the raw code and typing it in by hand (which is sadly open to typos and other errors).
The oflabaoslabaofly sequence I suggested was supposed to test forward left, forward right, and backward of the desired move. It obviously needs to be adjusted; oabaflabalabayflK seems closer to the mark: one step opposite the desired direction to find the edge, return, turn left 45°, return, turn left 90° (returning from front left, this turns it to front right), return, turn left 45° (returning from front right, this turns it to the desired direction), toggle range from step to slide, and go. This uses the edgefinder at the back of the move, but also checks the sides to make sure that there's no edge there as well. (It checks front and diagonally because directly to the side you'll find edges if you're trying to move diagonally from the corner.) Trying it on the Sandbox, it doesn't work, so I suspect that I missed something, but hopefully my description can help you find where I went wrong there.