implement-agent-teams

name	implement-agent-teams
description	Set up and validate Claude agent team configurations, collaboration flows, and operator guidance for team-based execution.

Implement Agent Teams

Implement team setup and coordination patterns for multi-agent workflows, including display modes and operational constraints.

Workflow

1. Validate complexity: Confirm team is needed (score 9-10). Can multiple specialized sub-agents work independently?
2. Define team objective: Clear goal requiring coordinated multi-agent execution
3. Identify members: Specialized agents with non-overlapping responsibilities
4. Set boundaries: Clear ownership and handoff contracts between agents
5. Configure metadata: Team metadata and task routing conventions
6. Plan coordination: Sequential vs parallel execution patterns
7. Document rules: Coordination rules and display mode choices
8. Evaluation strategy: How to test team-based workflows (20-50 scenarios)
9. Run checks: Environment and config validation
10. Verify fallbacks: Limitations and fallback behavior

Progressive Disclosure

Best Practices (Anthropic Guidelines)

• Core principles: ../../docs/ANTHROPIC_BEST_PRACTICES.md
  • Links to detailed references for complexity scoring, context management, and evaluation strategy

Team Design

• Team setup structure: references/team-setup.md
• Coordination contracts: references/team-coordination.md
• Display mode selection: references/team-display-modes.md
• Team best practices: references/team-best-practices.md
• Known limitations: references/team-limitations.md

Validation Tools

• Team config checker: scripts/check-team-config.sh
• Environment verifier: scripts/verify-team-env.sh

Templates

• Team config example: assets/templates/team-config-example.json
• Team task example: assets/templates/team-task-example.md

Related Skills

• Umbrella routing and component comparison: ../implement-claude-extensions/SKILL.md
• Sub-agent design: ../implement-sub-agents/SKILL.md

Sources

• https://code.claude.com/docs/en/agent-teams
• https://code.claude.com/docs/en/sub-agents