I (think) I have something similar to what you're trying to do. It's a module. I, like others, like to keep modules and binaries separate. In all my distributions that require a user usable binary, I just add it in so it gets installed. To do so, I do this in the Makefile.PL script:

my %WriteMakefileArgs = (
    NAME             => 'Addr::MyIP',
    AUTHOR           => q{Steve Bertrand <steveb@cpan.org>},
    VERSION_FROM     => 'lib/Addr/MyIP.pm',
    ABSTRACT_FROM    => 'lib/Addr/MyIP.pm',
    LICENSE          => 'artistic_2',
    EXE_FILES        => [qw(bin/myip)],
    ...
[download]

... see the last line? That will install the distribution's myip binary located in the distribution's bin directory into their path.

fwiw, I wrote that distribution (which fetches my devices public IP address) so that I can maintain VPN connections on the locations I have equipment even if they have dynamic/changing addresses.

I call that code from the update-ip binary in my Net::DNS::GoDaddy library, which then updates certain DNS A records for me through GoDaddy's API, so all of my locations are always up-to-date with their current public IP address. I can always use 'home.domain.com' for my home VPN connection etc.

Having the two files separate also makes it easier for others to hack at them, and for me to review the patches or pull requests. Documentation is separate, tests are separate, less chance I break one and not the other, and it doesn't have the feeling like there's executable code inside of a library. A library is to be read, not executed (imho), and executing a .pm file just seems so very wrong.