One reason I could see not sending a blanket uninstall command would be that you do not want the uninstall to occur if the Java 8 install has not yet happened.
Okay, so build a Patch Manager Computer Group from a report of systems with JRE8 installed and target the uninstall task to the PMCG contaning known (and confirmed) JRE8-based systems.
It also requires patch manager to have more network access and permissions over the client.
That it does. It also has the benefit of providing authoritative and immediate feedback that the installation actually occurred. If the motivation for doing this is "reducing attack profile", then I would think positive confirmation would be an important part of the task execution. Merely dropping an uninstall command in a PackageBoot executable does not provide that level of confirmation. (Nor does it preclude somebody else from reinstalling JRE7 the next day.), and it increases the risk that the JRE8 installation will FAIL, given that removing JRE7 is now a prerequisite to actually installing JRE8. Not to mention potentially leaving the machine with no JRE installed at all. Presumably the JRE is installed because it is actually NEEDED.
I have many machines that patch manager can not issue commands to
Patch Manager also has an optional agent that addresses this challenge.