the workaround i have at the moment is that audacity lets you export cue points as a .txt file.
As long as you don't add any labels to your cue points, then that text file contains a list of cue point times which you can read directly into an array in pd.
in audacity, do the following:
Project -> New Label Track
select your cue points (Add Label at Selection)
File -> Export Labels
however, a pd object to read the cue points from .wav or .aiff files
would be very useful. from what i have been reading, all the cue
points are stored as a single line of data in the header of the file.
here's a cue point reader as a supercollider patch:
http://www.create.ucsb.edu/pipermail/sc-users/2006-November/029666.html