Commands
Every command Maestro adds, plus the omp and ponytail commands it leans on.
Maestro adds two commands and one agent. Everything else on this page comes from omp or ponytail and is listed so you know where it lives.
/maestro
| Command | What it does |
|---|---|
/maestro | Shows the active ensemble — or custom if your roles don't match one exactly — and the choices. |
/maestro mix | Claude Opus orchestrates and plans, GPT Sol/Luna do the coding, Claude Sonnet does design and vision. The live session moves to the new lead. |
/maestro anthropic | Claude only: Opus thinks and plans, Sonnet codes, Haiku handles the small stuff. The live session moves to the new lead. |
/maestro openai | ChatGPT/Codex only: GPT-6.1 Sol thinks and codes, GPT-6 Luna handles the small stuff. The live session moves to the new lead. |
Tab completion offers the ensemble names with their descriptions. A successful switch confirms with:
Maestro: switched to "mix" — Claude Opus orchestrates and plans, GPT Sol/Luna do the coding, Claude Sonnet does design and vision/jev-route
| Command | What it does |
|---|---|
/jev-route | Same as status. |
/jev-route status | Enabled, verbose, key present, route counters, subagent counters and the last decision. |
/jev-route on | Turn routing on for this session. |
/jev-route off | Turn routing off for this session; a router-owned @smol turn returns to @default. |
/jev-route verbose | Show a notification for each main-turn routing decision. |
/jev-route quiet | Hide them again (the default). |
/jev-route test <prompt> | Dry-run the main-turn router on a prompt: Layer 0, Jev's scores, the route and why. |
/jev-route test-task <task> | Dry-run the subagent classifier: class, confidence, delegate probability and role. |
On and off last for the current session; the router starts enabled. A dry run looks like this (scores are an example):
/jev-route test Fix the typo in docs/getting-started.mdTEST: route=smol tier=economy/0.96 complexity=0.30/0.91 downgrade=0.95 impact=0.03 web=0.04 512ms — all conservative @smol downgrade gates passed/jev-route test Rotate the staging credentialsTEST: route=default 0ms — Layer 0: high-impact keyword guardsmart-commit agent
Maestro ships one agent definition, smart-commit. It runs the whole commit workflow on your @commit model with low thinking and only the read, bash and yield tools:
- Inspects the branch, upstream, status and diffs.
- Stops without committing on anything unsafe — merge conflicts, suspicious generated files, ambiguous unrelated changes, unusual submodules, unclear large binaries.
- Stages only the coherent logical change, reviews the staged diff, and writes a concise imperative subject (conventional commits if the repo uses them).
- Commits once and pushes — never amends, rewrites history, bypasses hooks or force-pushes.
- Reports the hash, subject, branch and push result.
Ask the lead to use it, or spawn it with omp's task tool.
From omp
/login | Log in to Anthropic and/or OpenAI Codex (ChatGPT plan). |
/model | Pick models; the Roles view changes any single role. |
/vibe | Director mode — the lead only reads and directs; fast workers use @smol, good workers @task. |
From ponytail
/ponytail switches on “lazy senior dev” mode from ponytail by Dietrich Gebert: the smallest complete change. See its repository for levels and options.