This might smack more of reinforcing fundamentals than deepening knowledge, but one thing that seems to work for me is to go through some more extensive shell scripts and see if you can't convert them into Perl, aiming at a reduction in code length and (hopefully) an increase in clarity. This might teach you something new about the shells as well.

Of course, this doesn't help you with broader design issues, and if you convert a poorly-thought-out script into perl, you get poorly-thought-out perl. Heck, even some well-thought out shell scripts might ignore perl-only idiomatic solutions, but you get the idea: use something that's been created as a proof-of-concept, and mirror that functionality in perl. Redesign if you deem it necessary.

Any *N*X box likely has a wealth of examples to draw from; and books on shell scripting should have shorter examples to copy. I'm sure googling will find you even more.

(This supplements the "write something to scratch your personal itch" line of advice I often hear; sometimes you don't even know what itches until you get further out into the forest.)

Something worth thinking about.