You've more or less described 1618's shop. Of note:
-There is no shame in using the kitbot. Get it together, make whatever cheap improvements you can (for us, that was going 6WD, riveting the frame, and adding AM Super Shifters), and focus on your manipulator.
-Think simple materials. We built the main arm for our 2007 robot, Uppercut, with a few pieces of angle aluminum and one length of PVC pipe from the local hardware store. Bolt it all together, put a window motor (or its equivalent for 2009) and you've got a single-stage arm. Now focus on a manipulator.
-Focus what machining resources you might be able to scrounge on the things you just can't do off-the shelf...but if you can, alter your design with these things in mind.
-Stephen Kowski pointed this one out to me: don't ask "What Would Beatty Do?", ask "What Would BeachBots Do?". Look at 330
the past four seasons. Together, those robots have won one four regionals (and finalist at two more), Curie in 2007, IRI in 2008, and the whole enchilada in 2005. They're all
dead simple; single joints for everything but 2006 (when an arm wasn't really an option). Build it as simple as possible, then refine that design as best you can.
More time, shorter letter.