orchestrator skill
A skill that turns the main Claude Code session into a dispatcher: tasks go to background subagents and the chat never blocks.
- 79 ★ on the repository
- Last commit
- MIT
Extended access
Why this level
- The skill launches several background subagents that edit project files and spend tokens
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add skills/olivieropinotti-orchestrator-skillDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- The skill launches several background subagents that edit project files and spend tokens
This is third-party code. Review the repository files before installing.
What it does
The `/orchestrator` skill hands every task to a background subagent. The main session only plans, dispatches, tracks and reports, so the user keeps talking while three to five agents work. Each message is split into independent tasks dispatched in parallel, and follow-ups go to the agent already on that topic so its context stays intact. A full brief is written for each agent: goal, facts, files it owns, what it must not do and what done means. The model is picked per task: Opus for judgment, Sonnet for tight checkable work. A small ledger `.claude/orchestrator.md` tracks every agent and survives context compaction.
Who it is for. People who work on several tasks at once in one Claude Code session and do not want to wait for the main agent to finish each one.
Good fit when
- Three independent edits at once in different parts of a project.
- During a long task you want to ask for status and give a new assignment.
- A clarification to a running task must reach the same agent with its context.
- A long-context session compacts and the list of running agents must be kept.
Not a fit when
- The task is single and short: dispatching only adds overhead.
- Strict control over every step is needed: the main session does not do the work itself.
- The token budget is limited: parallel agents spend it faster.
Example request
/orchestrator Fix the auth test, update the README and find the cause of the slow build.Limitations
The skill is written for Claude Code and relies on its background subagents. Three to five agents work at once and each spends tokens separately, so the bill grows faster than with a single session. Model selection per task is set in the skill text and actual cost depends on your subscription. The ledger `.claude/orchestrator.md` is created in the project and updated as work goes on. To leave the mode say "stop orchestrating". The repository is very small: one `SKILL.md` and an MIT license.
How to disable. Say "stop orchestrating" and delete the skill from `~/.claude/skills/`. Delete `.claude/orchestrator.md` in the project by hand.
FAQ
How do I turn the mode on and off?
Turn it on with `/orchestrator` and off with the phrase "stop orchestrating".
Which control commands exist?
`status` lists every agent and state, `stop <name>` or `stop all` stops them and `redirect <name> to ...` changes the task.
Where is the agent list stored?
In `.claude/orchestrator.md`. It survives context compaction.
How many agents work at once?
Three to five per the README.
How does the skill choose a model?
Opus for tasks needing judgment, Sonnet for tight tasks with a checkable result.
Discussion
No questions or reviews yet.
Related
SkillsOne agent launches others in the background: Claude Code, Codex, Copilot CLI, Grok Build and Pi, with a model chosen per stage.
A template sets up a multi-agent environment in a repository: Claude Code conducts Sonnet and Opus subagents and calls Codex CLI.
Multi-agent orchestration for Claude Code: teams, roles, autopilot and task routing across skills and subagents
A skill breaks a requirement into an Orca task graph, and a browser viewer shows it and runs workers along the dependencies.
A planner for Claude Code, Codex and OpenCode: one goal becomes a task plan and independent tasks run in parallel
Skills and the porch CLI for agent orchestration: a strong model plans, cheaper ones code, several model families review.
README in short
The README briefly explains the dispatcher idea, installation, capabilities and control commands. The skill also has a page on skills.sh.