The IRAF script MASKALIGN is used to align your masks. Set the COSMIC base position angle to the correct value, which is indicated in the slitmask plot label. First, take a short (30s) exposure of the field you will be using your mask on. Do not use any filter (filter position 1, clear), and leave the aperture (aperture position 1, clear) and grism (either position 2 or 4) wheels open.
When the image has read out, display it (with the [*,-*] orientation). Execute MASKALIGN, which will prompt you for the input file with slit positions (``mask.ap'', but without the .ap), and the image to work on. This program will then draw circles around the positions corresponding to the desired positions (in the slitmask) of the objects you have identified. This may be many arcseconds away from their actual positions in the image! Move to the objects in the order they were entered into the file, and at each one, put the cursor on the object, then hit the spacebar. After the last object, type ``q'' (for ``quit''). (You will need to ``quit'' twice, first in the display window, then in the IRAF window.)
This produces an output in the IRAF window of the mean offset for X and Y arc-sec and for theta in degrees, as well as errors from the mean fit in X and Y for each object. For a properly made and mounted mask, the errors of the objects with respect to the fit will be small, under 1.5 arc-sec, even when you are quite far off in pointing and/or rotation angle. If you see larger errors, either the coordinate list fed into the mask design program is incorrect or there has been a mistake in how the mask was mounted. If you see large errors of alternating sign, the most likely flaw is that the mask has been loaded upside down, and needs to be flipped with respect to the holder.
You will then be prompted to type in the COSMIC base position angle, so the offets in X and Y can be rotated into offsets in RA and Dec. Move the telescope as required. If the offset dRA is positive, move East. If dDEC is positive, move the telescope North. Add dTHETA to the COSMIC base position angle and rotate COSMIC as required. If the required rotation angle is less than 0.1,then you should not adjust it.
Two iterations will probably suffice for most fields given reasonable centerfield coordinates and telescope pointing. You should obtain a guide star after the first iteration (i.e. after the first telescope offset, and before starting the second image exposure), as this is an integral part of the final alignment. If it is necessary to make small moves after the first iteration, the guider can be placed in ``follow'' mode, the telescope moved, and guiding restarted.
You should now insert the appropriate mask and grism, and start taking your spectra. The entire alignment procedure should take about 12 minutes (6 of which are readout).