Command reference
Every mini-orchestrator command, grouped by what it does. Pick a command to see its usage and examples.
→ How mini adapts to your project
Running mini autonomously (auto / --yolo) is at your own risk — read the note before you start.
Project setup
- mini init /mini:init Creates a new mini-orchestrator project in the current directory.
- mini import-gsd /mini:import-gsd Imports an existing GSD project from .planning/ into mini (one-off).
- mini map /mini:map Regenerates the machine-readable project map for the agent.
- mini audit /mini:audit Scans the existing code into a .mini/codebase.md overview.
The phase loop
- mini next /mini:next Proposes what should come as the next phase of the project.
- mini discuss /mini:discuss Opens a session to discuss the phase intent before planning.
- mini plan /mini:plan Breaks the current phase down into concrete steps.
- mini do /mini:do Runs Claude Code on the current phase or step and writes a report.
- mini verify /mini:verify In-depth human UI/UX review of the phase before it is closed.
- mini done /mini:done Human verification — confirms it works and moves the state forward.
The autonomous loop
- mini auto /mini:auto Chains phases on its own (next → plan → do → done), pausing for human checks.
- mini stop mini stopCLI only Run from ANOTHER terminal during an auto run: drops a cooperative stop signal (.mini/STOP) so the loop finishes its current step and stops cleanly.
State & control
- mini status /mini:status Shows where the project currently stands across its phases.
- mini undo /mini:undo Reverts the last state change by one step.
- mini model /mini:model Sets the Claude model used for the project (per scope).
- mini todo /mini:todo Keeps an archive of future ideas and changes for the project.
- mini changelog /mini:changelog Shows the project's CHANGELOG.md changes (read-only).
- mini doctor /mini:doctor Quick health check of the project setup (read-only).
No command matches your filter.