That's pretty tidy, for sure. It will even work on a DOS command line. Maybe you'd be interested in shortening what follows?

I worked an absurdly long time on it, trying to understand how hashes of arrays work. It borrows heavily from something by jcwren and from the Cookbook:

  perl -e ' use strict; foreach (map {split} map {lc}<>){
  push @{ $ARGV[0] {glue => sort split // }}, $_};
  foreach (values %{$ARGV[0]}){my %seen=(); my @anagrams=
  grep{ ! $seen{$_}++ } @$_ ; print join (" " => @anagrams>1 ?
  @anagrams : next ), $/ }'
[download]

This doesn't filter out non-alpha stuff, but it does eliminate duplicates. It's not quite as flexible as yours, either, but it does take words from STDIN ( - it's too long for in Windows, but it can be shortened - I'd leave in strict, though. There's funny stuff going on here that could cause unpredict- able results if the wrong thing is put on the command-line).

It will also take a redirect from any file (not just lists).

So, it could be used as "perl -e '..etc...' < ${pathto}/wirds to find anagram in that dictionary (yeesh). No globbing.

Or, it can be invoked as "perl -e '..etc..'__Type Words__ followed by CTRL-D to exit.
I hope it's at least amusing.
mkmcconn