Having been foolish enough to say "Sure, we can do that" in response to the relatively complicated patch scenario my supervisor wanted us to implement, I can offer advice, if not code-- Our code is heavily dependent on our environment, and probably wouldn't make much sense. It's also fairly hideous, so I'd rather not incriminate myself. ;)
The biggest lesson I learned when trying to do patching with puppet, is if it's anything beyond "package { ensure => latest }", don't do it with puppet.
Puppet is very good at configuration management-- But not so good at process management, by which I mean, puppet isn't very good at making a sequence of events happen in the right order, at the right time.
Initially, I tried to manage the patch process directly with puppet, and it nearly broke my brain... and the end result wasn't terribly stable, or easy to debug. Now that I've become much better at Puppet, it might have turned out differently, but I still try to live by the rule that Puppet manages configurations, rather than processes.
So I use puppet to deliver the appropriate scripts and configuration files that I use for my patch process-- I have a Debian wrapper script, and a Red Hat wrapper script, both of which read configuration files-- I have a cron job or two which does the prep work for the patch cycle, and I keep the patch configuration data inside Hiera-- otherwise, all the patch "logic" takes place on the managed host, rather than the puppet server.