Skip to main content
The /orchestrate command coordinates multiple specialist agents to solve complex problems requiring different domain expertise. This is the most powerful workflow for full-stack features.

When to use this

Full-stack features

Tasks spanning frontend, backend, and database

Complex tasks

Multi-step tasks requiring coordination

Multiple domains

Tasks needing different specialist perspectives

Comprehensive reviews

Security, performance, and quality analysis

Critical requirement

MINIMUM 3 AGENTS: Orchestration requires at least 3 different specialist agents. Fewer agents = simple delegation, not orchestration.

Two-phase orchestration

Phase 1: Planning

1

Project planning

The project-planner agent creates docs/PLAN.md with:
  • Task breakdown
  • Agent assignments
  • Timeline estimates
2

Checkpoint: User approval

You must approve the plan before implementation starts.The system asks:
✅ Plan created: docs/PLAN.md

Do you approve? (Y/N)
- Y: Start implementation
- N: I'll revise the plan

Phase 2: Implementation (After Approval)

Multiple agents work in parallel:
Parallel GroupAgents
Foundationdatabase-architect, security-auditor
Corebackend-specialist, frontend-specialist
Polishtest-engineer, devops-engineer

Agent selection matrix

Task TypeRequired Agents (minimum 3)
Web Appfrontend-specialist, backend-specialist, test-engineer
APIbackend-specialist, security-auditor, test-engineer
UI/Designfrontend-specialist, seo-specialist, performance-optimizer
Databasedatabase-architect, backend-specialist, security-auditor
Full Stackproject-planner, frontend-specialist, backend-specialist, devops-engineer
Debugdebugger, explorer-agent, test-engineer
Securitysecurity-auditor, penetration-tester, devops-engineer

Available agents (17 total)

AgentDomainUse When
project-plannerPlanningTask breakdown, PLAN.md
explorer-agentDiscoveryCodebase mapping
frontend-specialistUI/UXReact, Vue, CSS, HTML
backend-specialistServerAPI, Node.js, Python
database-architectDataSQL, NoSQL, Schema
security-auditorSecurityVulnerabilities, Auth
penetration-testerSecurityActive testing
test-engineerTestingUnit, E2E, Coverage
devops-engineerOpsCI/CD, Docker, Deploy
mobile-developerMobileReact Native, Flutter
performance-optimizerSpeedLighthouse, Profiling
seo-specialistSEOMeta, Schema, Rankings
documentation-writerDocsREADME, API docs
debuggerDebugError analysis
game-developerGamesUnity, Godot

Output format

You’ll receive an orchestration report:
## 🎼 Orchestration Report

### Task
[Original task summary]

### Agents Invoked (MINIMUM 3)
| # | Agent | Focus Area | Status |
|---|-------|------------|--------|
| 1 | project-planner | Task breakdown | ✅ |
| 2 | frontend-specialist | UI implementation | ✅ |
| 3 | test-engineer | Verification scripts | ✅ |

### Verification Scripts Executed
- [x] security_scan.py → Pass
- [x] lint_runner.py → Pass

### Key Findings
1. **project-planner**: Created 5-phase implementation plan
2. **frontend-specialist**: Built responsive UI with dark mode
3. **test-engineer**: 95% test coverage achieved

### Deliverables
- [x] PLAN.md created
- [x] Code implemented
- [x] Tests passing
- [x] Scripts verified

### Summary
[One paragraph synthesis of all agent work]

Usage examples

/orchestrate build e-commerce platform
Coordinate frontend, backend, database, and testing agents
/orchestrate security audit and fixes
Run security-auditor, penetration-tester, and devops-engineer
/orchestrate optimize performance
Use performance-optimizer, frontend-specialist, and backend-specialist

Key principles

Context passing: Each agent receives full context including user decisions and previous agent work.
Verification required: The last agent must run validation scripts (security_scan.py, lint_runner.py).
User approval: Implementation Phase 2 only starts after you approve Phase 1 plan.
  • /plan - Planning-only mode (no implementation)
  • /create - Simpler single-domain creation
  • /test - Verify orchestrated work
  • /deploy - Deploy after orchestration

Build docs developers (and LLMs) love