This seems like a very basic feature. A task's "handoff logic" and it's possible linear (associative) "terms" (i.e., user-to-user, user-to-agent, agent-to-agent, and agent-to-user handoff relations, etc) is what defines a task as a "collaborative" task--as opposed to any other type of task, which have no such systems of logically (necessary) ordered invariants. On the other hand, the pattern of constraints: bottlenecks, blockers, dependencies (often sub-tasks themselves) and triggers, combined with the logic of handoffs constituting any given task, are the meaning of a task as such, and give it a definite and finite structure with a starting phase, a termination phase, and the phases between the start and termination phases aka a lifecycle. Therefore the specific meaning of a concrete task as 'this' individual collaborative task, as opposed to any other particular collaborative task, is that it can be repeated as the self-same task at different times but with the same necessary logic of handoffs (the CPA must know the value of the corporation before producing the final draft of last year's financial statements, but the business appraiser cannot finalize their valuation conclusion without finalized financial statements, and so there is a "handoff logic" where one professional takes the task as far as they can before they must finally "handoff" the task to another professional who take the task as far as possible, then turn around and deliver their results back to the first professional, finally allowing the two professionals to complete both of their high-level tasks by coming together and performing a final sub-task of 'reconciliation' which removes the final constraint/blocker and triggers the termination phase for the high-level task.

So yeah, I'm surprised this isn't already complete hahaha