formalization of @zhangdjr 's task
in accordance with the improvements made at https://github.com/seg-bio/segbio_prototype , we should ingest some of the work done there into the current agent implementation in the client.
very concretely, the agent should be able to—asking for approval first—be able to run commands and/or spawn tasks in addition to just giving heuristic advice as it does right now
continue specific discussion on this thread