creating-agent-teams

Creating Agent Teams

A Claude Code plugin with a skill for deciding when and how to create agent teams - whether a task needs a single agent, parallel subagents, or a coordinated team.

What This Plugin Does

This plugin provides a skill that helps you:

• Analyze tasks to determine the right approach (single agent vs parallel vs team)
• Select the right model tier (Haiku/Sonnet/Opus) for each role
• Choose agent types (Explore, general-purpose, code-reviewer, etc.)
• Compose effective teams with proper communication patterns
• Avoid common mistakes like token waste, scope creep, and coordination overhead

Prerequisites

Agent teams are experimental and disabled by default. Enable them in your Claude Code settings:

// ~/.claude/settings.json
{
  "env": {
    "CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS": "1"
  }
}

Official Documentation

For complete details on agent teams, see the official Claude Code documentation:

Agent Teams Documentation

Installation

Option 1: Plugin Install (Recommended)

Add this repository as a marketplace and install:

/plugin marketplace add ZoranSpirkovski/creating-agent-teams
/plugin install creating-agent-teams@ZoranSpirkovski-creating-agent-teams

Option 2: Clone to Plugins Directory

git clone https://github.com/ZoranSpirkovski/creating-agent-teams.git ~/.claude/plugins/creating-agent-teams

Option 3: Clone to Skills Directory (Legacy)

git clone https://github.com/ZoranSpirkovski/creating-agent-teams.git ~/.claude/skills/creating-agent-teams
cd ~/.claude/skills/creating-agent-teams
# Move skill files to root for legacy skill loading
mv skills/creating-agent-teams/* .
rmdir skills/creating-agent-teams skills

Plugin Structure

creating-agent-teams/
├── .claude-plugin/
│   └── plugin.json              # Plugin metadata
├── skills/
│   └── creating-agent-teams/
│       ├── SKILL.md             # Core skill - decision framework
│       ├── research.md          # Reference - model comparisons, patterns
│       └── agent-prompt-template.md  # Templates for spawning agents
└── README.md

Key Concepts

Agent Roster

Agent	Model	Type	Role
Orchestrator	opus	general-purpose	Coordinates, never implements
Explorer	haiku	Explore	Finds files, patterns, conventions
Frontend	sonnet	general-purpose	React/UI implementation
Backend	sonnet	general-purpose	Server-side logic
API	sonnet	general-purpose	Endpoints, contracts
Reviewer	sonnet	code-reviewer	Quality gate

Communication Modes

1. Hub-and-Spoke (default) - All communication flows through the Orchestrator
2. Discussion Mode - Agents debate peer-to-peer for investigation/architecture decisions

Model Selection

• Haiku: Search, grep, boilerplate, data gathering
• Sonnet: Implementation, debugging, code review (default)
• Opus: Architecture, ambiguous requirements, orchestration

Usage

The skill triggers when you:

• Ask to "create a team" or "set up agents"
• Mention "agent team" or "multi-agent"
• Request parallel work on a complex task

Example:

Create an agent team to implement the user authentication feature.
We need frontend login form, backend auth service, and API endpoints.

Contributing

To submit this plugin to the official Anthropic marketplace:

• Plugin Directory Submission Form

License

MIT