yes! sudocode manages agents at different levels of abstraction. You can steer agents at the `spec` level by building up a spec that contains your intent and requirements, then generate `issues` that capture the implementation details. However you aren't limited to that pattern. You can also just manage agents at the `issue` level to better manage their execution state (think something like a kanban board).

In general, sudocode aims to preserve some of this implementation/planning state (context-as-code) in your codebase so it isn't lost in your chat logs. Keeping this context in your repo makes it accessible to coding agents long after code changes were made, so it functions as a sort of episodic memory for your codebase.