I think that example on the xna site was more for like fog, for what your showing i think you will need a couple extra steps.
You would need to pass to the shader, 4x vector3's that describe each of the two planes.
These would denote
1) A positional point that lies on each plane.
2) The planes normalized surface direction.
(you will need two as shown in your image)
On the shader you would then have to perform the following steps. (in the same coordinate system)
1) Subtract the current pixel position from both passed plane positions (planePos - pixPos) you gave to the shader and normalize those results
2) Then dot each of those results to the normal directions you passed into the shader for each plane.
3) Check the resulting sign of both dot products to ensure they are within the boundrys.
if both resulting signs are positive you draw the pixel.
if they are not you clip the pixel.
Or you can probably just multiply the results of the dots and pass it to clip.
It's basically the same as a polyhedron to point collision test you will of course need a seperate shader to draw the quad image as shown in the second image.
If this isn't clear post back and ill see if i cant write out a basic shader for the steps.