You can get rid of eval (potential security risk) and even define some new tests like so:

use strict;
use warnings;

my %tests = (
#  key,        description,            anonymous test function
    -s    => { dsc => "size > 0",      ftest => sub { -s $_[0] } },
    -T    => { dsc => "type Text",     ftest => sub { -T $_[0] } },
    -B    => { dsc => "is Binary",     ftest => sub { -B $_[0] } },
    -r    => { dsc => "is readable",   ftest => sub { -r $_[0] } },
    -z    => { dsc => "is Zero bytes", ftest => sub { -z $_[0] } },
    lt100 => { dsc => "small file",    ftest => sub { -s $_[0] < 100 }
+ },
);

my $myfile = $0;

print "Tests for '$myfile':\n";

foreach my $t (keys %tests) {
  my $result = $tests{$t}{ftest}->( $myfile ) || 0;

  print $result ? 'Passed' : 'Failed';
  print " [$t]\t test ($tests{$t}{dsc})\t with result ($result)\n";
}
__END__
Tests for 'ftest.pl':
Passed [-r]      test (is readable)      with result (1)
Passed [-T]      test (type Text)        with result (1)
Failed [lt100]   test (small file)       with result (0)
Failed [-z]      test (is Zero bytes)    with result (0)
Failed [-B]      test (is Binary)        with result (0)
Passed [-s]      test (size > 0)         with result (780)
[download]

sub { -s $_[0] } is a little shorter than  sub { my $filename = shift; return -s $filename; } which is more comprehensible.