Team Builder

Automated Agent Team composition and deployment. Analyzes user requirements to propose optimal team structure, with teammates leveraging all available skills and agents across every scope.

Overview

This skill automates the process of creating Claude Code Agent Teams by:

• Analyzing user requirements to recommend optimal team composition
• Discovering all available skills/agents across 5 scopes (project, user, global, plugin marketplaces, plugin cache)
• Generating task dependency graphs with parallel/sequential flows
• Injecting relevant skills into teammate spawn prompts
• Supporting auto, template, and manual composition modes

Args

--agents

-a

-a "planner,system-architect"

--skills

-s

-s "tdd-workflow,security-review"

--template

-t

-t feature-dev

--name

-n

-n "auth-team"

--model

-m

-m adaptive

--size

--size 3

--lead

-l

-l planner

--dry-run

--dry-run

--auto

--auto

--delegate

-d

--delegate

--plan-approval

-p

-p

--display

--display split

Workflow

AUTO Mode (default)

When no

--template

--agents

1. Analyze Request: Detect domain keywords to identify the type of work
2. Discover Resources: Run

   python scripts/discover_resources.py --format json

   to catalog all available agents and skills across all scopes
3. Score Relevance: Match agents/skills to detected domain using the Domain Detection table below
4. Propose Team: Generate optimal team composition with task dependencies
5. Confirm: Present composition for user approval (skip with

   --auto

   )
6. Deploy: Execute TeamCreate -> TaskCreate (with blockedBy) -> Task spawn (with team_name) for each member

TEMPLATE Mode (

-t

specified)

1. Load Template: Read matching template from

   references/team-templates.md

2. Customize Tasks: Inject user's request into task descriptions
3. Override: Apply

   -a

   ,

   -s

   overrides if provided
4. Deploy: Execute team deployment sequence

MANUAL Mode (

-a

only)

1. Build Team: Create team from specified agent types
2. Route Skills: Match

   -s

   skills to appropriate roles
3. Generate Tasks: Auto-create tasks from user request
4. Deploy: Execute team deployment sequence

Available Templates

feature-dev

investigation

refactor

security-audit

frontend

full-stack

documentation

exploration

See

references/team-templates.md

Domain Detection

Skill Injection

Embed in each teammate's spawn prompt to make skills discoverable:

You are the {role-name} on team "{team-name}". Your task is: {task-description}

## Available Skills
Invoke the following skills as needed during your work:
- /tdd-workflow: Test-driven development with 80%+ coverage
- /security-review: Comprehensive security checklist
- /code-reviewer: Code quality review
- /{plugin-skill}: {description from catalog}

## On Completion
1. Use TaskUpdate to mark your task as completed
2. Send a summary of your findings to team-lead via SendMessage

Skills listed are dynamically populated from

discover_resources.py

Model Selection Strategy

-m

deep

adaptive

fast

budget

Deployment Sequence

1. TeamCreate(team_name, description)
2. For each member:
   a. TaskCreate(subject, description, activeForm)
3. Set dependencies:
   a. TaskUpdate(taskId, addBlockedBy: [...])
4. For each member:
   a. Task(subagent_type, team_name, name, prompt, model, mode)
5. Assign initial tasks:
   a. TaskUpdate(taskId, owner: member-name, status: "in_progress")

When

--plan-approval

mode: "plan"

When

--delegate

Usage Examples

# Auto: compose from request analysis
/team-builder "Design and implement JWT authentication system"

# Template: use predefined template
/team-builder -t feature-dev "User management API"

# Manual: specify agents and skills
/team-builder -a "planner,frontend-architect,e2e-runner" -s "senior-frontend,e2e,ai-sdk"

# Dry-run: preview composition
/team-builder --dry-run "Large-scale refactoring"

# Budget + auto: cost-optimized auto-deploy
/team-builder -m budget --auto "Documentation cleanup"

# With plan approval for high-risk work
/team-builder -t full-stack -p "Payment processing system"

# Delegate mode: lead coordinates only
/team-builder -d -t feature-dev "New notification system"

# Split display for real-time monitoring
/team-builder --display split -t exploration "Evaluate microservices migration"

Resources

scripts/

• discover_resources.py

  : Discovers all available agents and skills across all 5 scopes. Run with

  --format json|table

  and

  --scope project|user|global|plugin|all

  .

references/

• team-templates.md

  : 8 predefined team templates with member definitions, task flows, and spawn prompt templates.

• composition-guide.md

  : Team composition best practices including sizing, model selection, task granularity, file conflict avoidance, and anti-patterns.