Confusion regarding agent responsibilities and delegation #97
Description
I see a lot of gsd-XXX agents named by function (see screenshot). However, the behavior seems inconsistent. Sometimes agents trigger each other, and sometimes they perform tasks outside their implied scope—for example, gsd-planner modifying code directly instead of calling gsd-executor.
Unlike projects like oh-my-opencode that have a clear hierarchy with a single orchestrator, the boundaries here are confusing. Could you clarify when it is proper to manually trigger /gsd-XXX agents versus letting the system handle the delegation?
