Guide · Agent skill
How to use the Orchestration Team Coordinator agent skill
Updated · written with AI from the project's own documentation and checked against our catalogue data
The orchestration-team skill tells an AI agent to act as the coordinator of other AI agents working on a shared project, instead of writing the implementation itself. It is written for people running OpenRig multi-agent rigs who want an orchestrator to break work into assignments, dispatch them, and watch for agents that finished or got stuck. By the end of this guide you will know what the skill changes, how to install it, how to ask for it, and where it does not fit.
Orchestration Team Coordinatormvschwarz/openrig · Apache-2.0 Download the skill Quick start
- 1Download the ZIP. Download the orchestration-team skill ZIP from this page.
- 2Unzip into a skills folder. Unzip into ~/.claude/skills/ for personal use, or into a project's .claude/skills/ folder.
- 3Or upload in the Claude apps. In the Claude apps, open Settings → Capabilities → Skills and upload the ZIP.
- 4Confirm the OpenRig CLI works. Run rig whoami --json from your rig to check the CLI can identify your seat.
- 5Start your rig with context files. Make sure project.yaml, mission.yaml and slice.yaml exist for the work you want coordinated.
- 6Ask for coordination in plain language. Request assignment breakdown, dispatch, or blocked-work checks from the coordinator seat with a clear stop condition.
What the orchestration-team skill makes the agent do
The skill is a single SKILL.md instruction file. When loaded, it tells the agent that its job is to keep the team productive rather than to do the implementation work. The agent reads its project, mission and slice context, breaks work into clear assignments, dispatches implementation, design, QA and review work, and watches for idle agents, blocked agents and coordination gaps.
Monitoring is kept light by design. The skill prefers reading status from the queue and a periodic watchdog wake over constantly polling every agent. When an agent finishes without handing off or gets stuck, the skill tells the coordinator to correct that agent by teaching the handoff protocol, not by silently doing the agent's work.
- Breaks work into assignments and dispatches them to other seats
- Reads queue and seat status rather than capturing every pane
- Catches agents that went idle without a handoff
- Corrects stalled agents by re-teaching the protocol
Install the orchestration-team skill
Download the skill ZIP from this page. Unzip it into your personal skills folder, ~/.claude/skills/, or into a project's .claude/skills/ folder if you want the skill to apply only to that project. The skill's files, including SKILL.md, should end up inside a folder named for the skill.
In the Claude apps you can instead upload the ZIP under Settings → Capabilities → Skills.
- Personal install: ~/.claude/skills/
- Project install: .claude/skills/ inside your project
- Claude apps: Settings → Capabilities → Skills, then upload the ZIP
Requirements before you trigger it
The skill is built around the OpenRig CLI. Its instructions depend on the rig command with whoami, queue, ps and capture subcommands, so the CLI must be installed. You also need a running OpenRig rig that has project.yaml, mission.yaml and slice.yaml context files, because the skill resolves work from those files.
Examples in the skill pipe JSON output through jq for filtering, so jq is useful to have installed. The skill does not run scripts and does not use the network itself.
- OpenRig CLI (rig command) with whoami, queue, ps and capture
- A running OpenRig rig with project.yaml, mission.yaml and slice.yaml
- jq, referenced in filtering examples
rig whoami --jsonTrigger the skill with natural requests
The skill's description says it applies when you are coordinating assignments, selected review boundaries, or blocked work across a rig. Ask for those things in plain language from the coordinator seat. The agent will start with a startup sequence that runs rig whoami and then resolves your project, mission and slice context before dispatching anything.
Be specific about the outcome you want and the stop condition. The skill asks the coordinator to dispatch the smallest complete outcome, so a request with a clear finish line works better than an open-ended one.
- "Break this slice into assignments and dispatch implementation and QA to the right seats."
- "Check the queue for any seat that finished without handing off."
- "Coordinate the review boundary for this wave, then stop once review is queued."
- "Find the blocked work across the rig and tell me what each seat needs."
What the coordinator does on each check
The skill describes a cheap sweep. The coordinator reads the queue and a filtered view of the rig's seats first, and only captures the last few lines of a seat that looks idle or suspicious. If the owner is active, the coordinator does nothing.
The skill also warns against reading more than a decision needs. Filter output to the rig or item in question, and do not pull whole-fleet status to answer a narrow question.
- Read the queue and filtered seat status first
- Capture pane output only for one named suspicious seat
- Take no action when the owner is active
- Narrow any command that returns unexpectedly large output
rig queue show <qitem> --jsonSafety notes: scripts and network
The skill ships as instructions only. It runs no scripts and does not use the network. Its effect comes from the rig commands the agent runs under your account, such as dispatching work through the queue and reading seat status. Review what the coordinator dispatches, especially on shared or production-adjacent rigs.
The skill notes that continuous monitoring burns shared account capacity. Its guidance is to keep monitoring bounded to a named risk and to end close watching once the condition clears.
- No scripts included in the package
- No network use declared by the skill
- Actions run through your OpenRig CLI and its queue
- Keep close monitoring bounded and time-limited
What people use it for
- Splitting a slice into implementation, design, QA and review assignments across seats in one rig
- Checking the queue for agents that finished a task but never handed it off to the next seat
- Coordinating a review boundary for a wave of work and stopping once review is queued
- Finding blocked work across a rig and reporting what each blocked seat needs
Frequently asked questions
How do I install the orchestration-team skill?
Download the skill ZIP from this page and unzip it into ~/.claude/skills/ for personal use or into a project's .claude/skills/ folder. In the Claude apps, you can upload the ZIP under Settings → Capabilities → Skills instead.
Is the orchestration-team skill free?
The source material lists the licence as Apache-2.0. Check the repository's licence file for the full terms before you use or redistribute it. This page does not state any price.
What do I need to run the orchestration-team skill?
You need the OpenRig CLI with the rig whoami, queue, ps and capture subcommands, plus a running OpenRig rig that has project.yaml, mission.yaml and slice.yaml context files. jq is referenced in examples for filtering output.
Does the skill run scripts or use the network?
According to its listing, the skill runs no scripts and uses no network. Its actions come from the rig commands the agent runs through your OpenRig CLI.
Can I use this skill without OpenRig?
The instructions depend on OpenRig rig commands such as rig whoami, rig queue and rig ps, so the skill is not designed to work without the OpenRig CLI. Single-agent or simple one-off tasks do not need it.
Is this a general-purpose project manager skill?
No. The skill assumes a specific queue-based handoff protocol defined elsewhere in the OpenRig project. If you want a general task tracker, pick a tool built for that purpose.
New agent skills, repos and AI tools — every week
The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.
Our verdict
Use this skill if you run OpenRig multi-agent rigs and want an orchestrator agent that routes work between seats with a documented, light-touch monitoring routine. Skip it for single-agent work, simple one-off tasks, or any setup without the OpenRig CLI, because the instructions will not match your environment.
Download the skill