foreach my $t (keys %tests) {
    my $cmd="$t '$myfile'";
    print "running [$cmd]\n";
    if ( eval{"$cmd"} )
[download]

running [-r './test.pl']
 Passed [-r] test, ./test.pl is (is readable)
running [-T './test.pl']
 Passed [-T] test, ./test.pl is (type Text)
running [-z './test.pl']
 Passed [-z] test, ./test.pl is (is Zero bytes)
running [-B './test.pl']
 Passed [-B] test, ./test.pl is (is Binary)
running [-s './test.pl']
 Passed [-s] test, ./test.pl is (more than 0 bytes)
[download]

eval{"$cmd"}

That's true if "$cmd" is true. If you want to evaluate $cmd, then write:

eval $cmd
[download]

Great catch JavaFan!

You can also get contradictory answers, in this case, both -s and -z returning false, if ./test.pl does not exist. It clearly isn't your problem here, but if you are applying this sequence of tests to arbitrary files, you might want to do the test for existence separately before passing the file through the gauntlet of your hashed up tests.