So first off, in general I think that if it works, it's good. Which is not to say it couldn't be better; but is fixing this worth your time?

GrandFather's right that it would be more natural to just spin up a new process upon need, rather than relying on something sitting in memory chewing up resources. Auxiliary concerns include do you need your "mighty agent" to have persistent memory? Do requested tasks need to be performed sequentially? Does ordering of those tasks matter? Are there security/permissions aspects to consider (i.e. does your "mighty agent" need greater permissions than your other agents?).