Has anyone tried the chew back-anneal in vitro DNA assembly method (known as Gibson Assembly) for difficult sequences, like GC-rich sequences? How big constructs could you efficiently assemble? Did you use the original protocol or some optimizations?
You could try running your difficult sequences through their tool to see what it suggests for primers. It apparently gives you a "value for the Gibbs free energy of the worst case secondary structure of the primer."
I also haven't tried it, but theoretically GC-rich sequences will interfere with Gibson assembly. Since unique homology between the ends is key to the assembly process, repetitive sequences which cause non-unique overhangs will increase the possibility of incorrectly ordered assembly. One solution would be to use alternative codons where possible to decrease repetitiveness.
I'm sure all the overlap design tools will try to avoid repetitive overlaps. Prof Jim Hasselhoff at Cambridge Plant Science is a big proponent of Gibson assembly, it was one of his students who wrote Gibthon. If you email him I'm sure he'll give you specific advice.