Delegate the work to a Team

Goal: when multiple projects / roles need to advance in parallel, let the Rind Team watch them for you instead of switching windows all by yourself.

Open the management page

With the input box empty, press ← to enter Agents Management. The Manager is on this page too, and you can use natural language to have it help you put a team together and hand out work.

The other route is the rind agents command in a terminal:

rind agents team create product
rind agents team add product /work/project --position Builder
rind agents team add product /work/reviewer --position Reviewer
rind agents task product reviewer "Review the current change and report evidence."

Mental model

  • Every member is an ordinary folder: the folder's basename is its initial name, and --position is only a display label.
  • A Team is a registration relationship: the same directory can join multiple teams; when reusing a directory across teams you will be advised to make a copy first, and --share is the explicit shared option.
  • Shared directories write serially; worktrees can run in parallel: rind agents team worktree product <name> <repo> <branch> opens a separate member for parallel development.
  • The first member is only the initial leader, and it gains no extra file permissions for that.

Handing out work and getting it back

State the deliverable and the evidence when assigning a task; once a member finishes, a receipt wakes the person who assigned it. A task only counts as done with a valid receipt + a confirmed end of execution; a run that lost contact midway is marked Unconfirmed and is not automatically re-run.

A blocker must name the responder (user or a member) and the required action, otherwise the person who assigned the work gets nothing back.

To watch or step into a specific member's session: enter its conversation from the management page, and it is no different from an ordinary Rind session.

Finer-grained rules

For adding and removing members, hierarchy adjustments, importing old Teams and recovery from failures, see the Agents Management documentation.

Back to the scenario index