| name | rhizome-development |
| description | Describes the current local development setup, for when we want to perform development tasks on the actual Rhizome app |
The preferred setup is a CMUX workspace (expect us to not work on more than one feature simultanously on a single machine,
so we also confine ourselves to a single workspace and you can expect us to not need to reach outside our single cmux workspace), where we, in the left pane of two panes, coordinate from a claude session (the coordinator), in the upper pane of a vertical split, in its terminal surface, another agent, which sits in the lower pane in its
vertical split, in its own terminal surface.
That agent is the doer. We want the doing be happening in a sandbox.
Normally in one where the internet is cut off (make yolo), and when the feature(s) we are working on demand,
one which has vector db support. For reference, see rhizome/README.md.
Interaction happens between the human and the coordinator, and then, autonomously, between the coordinator and the doer.
I want full visibility, that's why I want handoffs to happen in the cmux surfaces (use cmux CLI).
We commit directly to the main branch, and want specifically most, if not all, commits, come from the in the box agent,
as it has its own git handle which we can easily recognise.