How many agents can I spawn with the Team skill?

You can spawn between 1 and 20 agents. The skill can also auto-size the team based on the complexity and decomposition of the task.

What happens if a task fails during execution?

The skill includes a dedicated 'team-fix' stage that automatically generates fix tasks based on verification failures and loops until the criteria are met or a limit is reached.

Can I use different models for different agents?

Yes, the orchestrator automatically routes tasks to specialized agents like 'architect' (Opus) or 'executor' (Sonnet) based on the project stage and task characteristics.

How does the lead agent maintain context across stages?

It uses a lightweight handoff system that writes key decisions, risks, and rejected alternatives to Markdown files, ensuring context survives agent restarts or session resumes.

What is the difference between this and the legacy /swarm skill?

This skill utilizes native Claude Code team tools, replacing external SQLite dependencies with built-in team management, inter-agent messaging, and task dependencies.

Multi-Agent Team Orchestrator

Name: Multi-Agent Team Orchestrator
Author: Yeachan-Heo

byYeachan-Heo

•

7,661

•

协作工具

Coordinates multiple AI agents to execute complex, multi-step tasks using native Claude Code team management and staged pipelines.

The Team skill enables sophisticated multi-agent orchestration within Claude Code, allowing users to spawn up to 20 coordinated agents to tackle large-scale software projects. It automatically decomposes high-level tasks into manageable subtasks, routes specialized agents—such as architects, designers, and testers—to specific stages, and manages inter-agent communication and task dependencies. With a robust staged pipeline covering planning, execution, verification, and automated fixing, it ensures high-quality outcomes for complex engineering efforts like project-wide refactoring or building full-stack features.

主要功能

01Intelligent agent routing based on task type (architects, executors, testers)

02Persistent stage handoffs to maintain context across agent restarts

03Automatic task decomposition with dependency and ownership management

04Native multi-agent orchestration for up to 20 coordinated workers

05Staged execution pipeline: planning, PRD, execution, verification, and fixing

067,661 GitHub stars

使用场景

01Building complex features requiring simultaneous backend, frontend, and testing work

02Refactoring large codebases or fixing project-wide TypeScript and linting errors

03Conducting comprehensive security audits and architectural reviews across multiple modules

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add yeachan-heo/oh-my-claudecode team

For use in Claude.ai and ChatGPT

Download Skill