| name | team |
| description | Multi-agent coordinated execution |
| argument-hint | <N:agent> <task description> |
| level | 4 |
Team orchestrates multiple specialized agents to work on a task in parallel. Each agent handles a different aspect, then results are synthesized into a cohesive solution.
<Use_When>
- User says "team", "multiple agents", "coordinate"
- Complex task with distinct domains (frontend + backend + tests)
- User wants parallel expertise (security + performance + functionality)
- Large feature requiring parallel implementation streams
</Use_When>
<Do_Not_Use_When>
- Simple single-domain task
- Tasks requiring sequential thinking
- Quick fixes
</Do_Not_Use_When>
<Execution_Policy>
- Assign agents to distinct, non-overlapping responsibilities
- Agents work in parallel on their domains
- Synthesize results into unified solution
- Run integration tests after synthesis
</Execution_Policy>
1. **Analyze**: Break task into domains (frontend, backend, tests, docs, etc.)
2. **Show Team**: Present the agent assignments
- List: "Frontend → @ui-engineer, Backend → @api-designer, Tests → @test-writer"
- Ask: "Does this team composition look right?"
- **CONFIRMATION REQUIRED before proceeding**
3. **Execute**: Run agents in parallel
4. **Synthesize**: Combine outputs into cohesive solution
5. **Integrate**: Ensure parts work together
6. **Test**: Run full test suite
<Agent_Assignments>
| Domain | Agent |
|---|
| Architecture | @architect |
| Frontend | @ui-engineer, @style-expert |
| Backend | @api-designer, @db-engineer |
| Testing | @test-writer |
| Security | @security-reviewer |
| Performance | @perf-optimizer |
| Documentation | @doc-writer |
| </Agent_Assignments> | |
User: "team 3 @executor @test-writer @code-reviewer build user authentication"
Why good: Clear agent assignments, parallel workstreams
User: "team @architect design the system @backend implement API @frontend build UI"
Why good: Sequential pipeline with domain specialization
User: "team 5 agents fix a typo"
Why bad: Overkill for simple task
