Team Assemble

Dynamically assemble expert agent teams for complex tasks using Claude Code's agent teams feature.

What It Does

Instead of manually designing and coordinating multiple agents, team-assemble:

1. Analyzes your task and identifies relevant codebase areas
2. Scouts the codebase to understand what agents are needed
3. Designs an optimal team with the right roles and dependencies
4. Executes agents in parallel where possible
5. Validates results against acceptance criteria
6. Cleans up the team automatically

Prerequisites

Agent teams are experimental and must be enabled first. Add to your settings.json:

{
  "env": {
    "CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS": "1"
  }
}

See skills/team-assemble/references/enable-agent-teams.md for detailed setup instructions.

Usage

Tell Claude what you need and ask it to assemble a team:

Assemble a team to refactor our authentication from session-based to JWT

Use team-assemble to evaluate caching strategies — Redis vs Memcached vs in-memory

Assemble a team to extract shared utilities from three microservices into a common library

The skill will guide you through approval gates before execution begins.

How It Works

6-Phase Workflow

Phase 1 → Phase 2 → Phase 3 → Phase 4 → Phase 5 → Phase 6
Task       Codebase   Integrate  Execute   Validate   Complete
Analysis   Scouts     & Confirm                       & Cleanup

Model 3-Tier Strategy

Agents are assigned models based on their role:

Key Features

• Dynamic agent design — no fixed catalog, agents are tailored to each task
• Parallel execution — independent agents run simultaneously
• Acceptance criteria — every team has measurable validation criteria
• Verify/fix loop — QA validates, support fixes (max 3 rounds)
• Two approval gates — user confirms scope (Phase 1) and team composition (Phase 3)

Plugin Contents

team-assemble/
├── .claude-plugin/
│   └── plugin.json
├── skills/
│   └── team-assemble/
│       ├── SKILL.md                        # Main skill definition
│       └── references/
│           ├── agents.md                   # Agent example bank
│           ├── prompt-templates.md         # Prompt templates for all phases
│           ├── examples.md                 # Worked examples
│           └── enable-agent-teams.md       # Setup guide
└── README.md