Reference
Commands
| Command | What it does |
|---|---|
mmt | home — list your teams with [local]/[global] scope |
mmt list teams|agents|skills | list what you have (aliases: mmt teams · mmt skills · mmt agents) |
mmt run <team> "task" | run on your subscription (opens a Claude Code session) |
mmt new team|agent|skill <name> [text] | create one — bare = describe it, Claude drafts it; --ui = author it yourself |
mmt edit team|agent|skill <name> ["change"] | update one — bare = describe the change; --ui = edit it yourself |
mmt show team|agent|skill <name> | print one |
mmt delete team|agent|skill <name> | delete one (alias: rm) |
mmt export <team> [dir] | write a reviewable bundle: team.yaml + agents/ + skills/ (--force to overwrite) |
mmt import <dir> | install a bundle after a manifest + consent (or legacy mmt import '<token>' / <file>) |
mmt help | list everything |
run flags
| Flag | Effect |
|---|---|
--model opus|sonnet|haiku | override the model for the whole run |
--sim | offline simulated preview (no Claude calls, no cost) |
Team file format
A team is <name>.team.yaml:
team: spec-to-prod # name (matches the file)
about: from a task to a reviewed PR
model: sonnet # optional default model for all members
lead: > # optional custom lead instructions
You are the topic owner. Keep the team focused and ship.
steps: # ordered
- member: strategist
does: turn the task into a short plan
- member: coder
does: implement and open a PR
skills: [github-pr, ticket-status]
model: opus # optional per-member model
- loop:
until: reviewer approves
max_rounds: 3
steps:
- member: reviewer
skills: [github-comment]
- member: coder
skills: [github-pr]
- member: qa
skills: [run-tests]
- member: reporter # optional — posts an audit comment on the PR
skills: [publish-report]
Member fields
| Field | Meaning |
|---|---|
member | the role name |
does | one line describing its job |
skills | list of skills it plugs in |
model | opus / sonnet / haiku (optional) |
Loop fields
| Field | Meaning |
|---|---|
until | exit condition; also names the gate member |
max_rounds | hard cap on iterations |
steps | the members that repeat (first = gate) |
Team-level
| Field | Meaning |
|---|---|
team | name |
about | one-line description |
model | default model for members |
lead | custom team-lead instructions (optional) |
report: github | opt-in: post a run audit to the PR |
Skill file format
A skill is <name>/SKILL.md:
---
name: run-tests
description: Run the test suite, report pass/fail, surface failing cases.
---
# run-tests
Run the project's tests, summarize pass/fail, and list any failing cases.
Files & locations
| Path | What |
|---|---|
~/.my-mini-team/teams/ | your global teams |
./teams/ | project-local teams (--local) |
~/.my-mini-team/skills/ | your mmt skill library |
~/.my-mini-team/runs/ | full run reports |