orchestrating-swarms

name	orchestrating-swarms
description	Master multi-agent orchestration using Claude Code's swarm system. Use when coordinating multiple agents, running parallel code reviews, creating pipeline workflows with dependencies, building self-organizing task queues, or any task benefiting from divide-and-conquer patterns. This facade loads specialist skills for primitives, spawning, operations, and patterns.

Claude Code Swarm Orchestration

Master multi-agent orchestration using Claude Code's TeamCreate, SendMessage, TeamDelete, and Task tools.

This skill is a facade that routes to 4 specialist skills. Load whichever you need.

---

Specialist Skills

Core Concepts

Skill(command: "swarm-primitives")

What teams, teammates, tasks, and inboxes are. File layouts, team config structure, lifecycle diagrams, message flow sequences.

Spawning Agents

Skill(command: "swarm-spawning")

How to create agents -- subagent vs teammate decision, built-in agent types (Explore, Plan, general-purpose, Bash, etc.), plugin agent types, spawn backends (in-process, tmux, iterm2), environment variables.

API Reference

Skill(command: "swarm-operations")

Tool signatures and message schemas -- TeamCreate, SendMessage (direct, broadcast, shutdown, plan approval), TeamDelete, Task tool parameters. Error handling, graceful shutdown sequence, crashed teammate recovery, debugging.

Patterns and Recipes

Skill(command: "swarm-patterns")

6 orchestration patterns (parallel specialists, pipeline, swarm, research+implement, plan approval, coordinated refactoring), 3 complete workflows, best practices, quick reference.

---

Quick Start

flowchart TD
    Start([What do you need?]) --> Q1{Your goal?}
    Q1 -->|Understand how swarms work| P[swarm-primitives]
    Q1 -->|Choose agent types or backends| S[swarm-spawning]
    Q1 -->|Look up tool API or message format| O[swarm-operations]
    Q1 -->|Build a workflow from a recipe| R[swarm-patterns]
    Q1 -->|Full reference on everything| All[Load all 4 skills]

---

Minimal Example

// 1. Create team
TeamCreate({ team_name: "my-team" })

// 2. Create tasks
TaskCreate({ subject: "Review auth", description: "Review auth module", activeForm: "Reviewing..." })
TaskCreate({ subject: "Review API", description: "Review API endpoints", activeForm: "Reviewing..." })

// 3. Spawn teammates
Task({ team_name: "my-team", name: "reviewer-1", subagent_type: "general-purpose", prompt: "Claim task #1, review it, send findings to team-lead.", run_in_background: true })
Task({ team_name: "my-team", name: "reviewer-2", subagent_type: "general-purpose", prompt: "Claim task #2, review it, send findings to team-lead.", run_in_background: true })

// 4. Collect results (messages arrive automatically)

// 5. Shutdown and cleanup
SendMessage({ type: "shutdown_request", recipient: "reviewer-1", content:  })
({ : , : , :  })

()

---

SOURCE: Claude Code v2.1.45 -- verified 2026-02-18