orchestrator

# Orchestrator Protocol

Multi-agent task orchestration with intelligent decomposition and parallel execution.

## Activation Banner (MANDATORY)

When skill activates, display:

```
(ORCHESTRATOR) ════════════════════════════════════════
  Mode: [/orchestrate | /parallel | /plan-only | auto]
  Complexity: [SIMPLE | MODERATE | COMPLEX | EPIC]
  Workstreams detected: [N]
════════════════════════════════════════════════════════
```

## Phase 0: Complexity Assessment

Before anything, assess task complexity:

| Indicator | Points |
|-----------|--------|
| Multiple distinct deliverables | +1 |
| Cross-cutting concerns | +1 |
| >500 lines affected estimate | +1 |
| Multiple file types/domains | +1 |
| Integration requirements | +1 |

**Complexity levels:**
- 0-1: SIMPLE - No orchestration needed
- 2-3: MODERATE - Light orchestration, 2-3 workstreams
- 4-5: COMPLEX - Full orchestration, 3-5 workstreams
- 6+: EPIC - Multi-phase orchestration, 5+ workstreams

**Decision:**
- SIMPLE: Respond normally, no orchestration
- MODERATE+: Activate orchestration protocol

## Phase 1: Task Decomposition

Break the task into atomic workstreams using the AND/OR model:

### AND Tasks (must all complete)
Tasks connected by "and", "also", "plus":
- "Add auth AND logging AND tests" -> 3 workstreams

### Sequential Dependencies
Tasks where one depends on another:
- "Create schema THEN implement API THEN add tests"

### Parallel Opportunities
Independent tasks that can run simultaneously:
- Architecture exploration (Explore agent)
- Pattern research (Explore agent)
- Dependency check (Explore agent)

### Decomposition Output Format

```
WORKSTREAM MAP
══════════════════════════════════════

Phase 1 (Parallel):
  [1] Research: Explore existing patterns
      Agent: Explore | Est: 2min
  [2] Research: Check dependencies
      Agent: Explore | Est: 1min

Phase 2 (Sequential):
  [3] Implement: Core feature
      Agent: general-purpose | Est: 10min
      Depends: [1], [2]

Phase 3 (Parallel):
  [4