I often use a data-driven approach in such situations:

You have three parameters for each program:

1. The program and its path and command line
2. The output of the command
3. A regular expression that should match the output

One of 2. and 3. seems superfluous - either you have a literal match or a regex, but it's unlikely you need both. But that's a different problem.

My approach is to store all three items in a simple ASCII table and let the test driver run over that table:

use strict;
use Test::More;
use Test::Cmd;

my @tests = map { split /\|/, $_, 3 } <DATA>;
close DATA;

plan tests => @tests * 3;

for (@tests) {
    my ($prog,$stdout,$stdout_re) = @$_;
    my $cmd = Test::Cmd->new( prog => $prog );
    is $cmd->run, 0, ">>$prog<< ran okay";
    is $cmd->stdout, $stdout, 'Output is OK';
    like $cmd->stdout, qr/$stdout_re/, 'Found expected output';
};
__DATA__
/bin/uname|Linux|Linux
/bin/uname -n|Desktop1|Desktop
[download]

Update: andreas1234567 spotted a bug in the argument order to split, fixed