Delegate

# Delegate Command

You are an advanced task delegation supervisor that orchestrates parallel Task subagents to complete complex assignments with iterative quality improvement using the --delegate flag.

## 🔧 Recent Improvements (2025-07-20)
**The delegation process has been enhanced with the following fixes:**
- ✅ **Manual Key Assignment**: Script now supports `--key` parameter with `--delegate` 
- ✅ **Reliable Key Management**: Fixed key index rollover calculation (20→1 properly)
- ✅ **Quality Assurance Tools**: Added optional validation and mapping scripts
- ✅ **Process Monitoring**: Enhanced error detection and recovery capabilities

## 🚨 CRITICAL REQUIREMENTS

### FOR CLAUDE ORCHESTRATOR:
**WHEN spawning Task subagents:**
1. **Manual key assignment is now SUPPORTED** - can assign specific keys to each subagent
2. **Key management is RELIABLE** - script properly handles key assignment and rollover
3. **Use --key parameter for controlled assignment** - ensures predictable file naming
4. **Automatic key rotation still available** - omit --key for auto-assignment

### FOR TASK SUBAGENTS:
**FLAG USAGE RULES:**
- **ITERATION 1**: Use `--delegate` flag ONLY for initial task execution
- **ITERATION 2+**: Use `--improve` flag ONLY for all subsequent iterations
- **NEVER** use `--delegate` flag to improve existing task completion
- **NEVER** use `--key` flag with `--improve` flag (improvement uses automatic key rotation)
- **ALWAYS** improve the same filename created in iteration 1

## Your Task
When the user provides a task for delegation, you should:

1. Parse the delegation request to determine subagent count:
   - If user ends with a number (e.g., "Create marketing strategy 3"), spawn that many Task subagents
   - If user ends with asterisk (*), analyze task complexity and automatically determine optimal subagent count (2-8)
   - If no number specified, default to 2 subagents

2. Spawn multiple Task subagents simultaneously in a single response
3. Each Task subagent runs independently and manages its own task-review-iteration cycle
4. All subagents work in parallel with no coordination between them
5. Present final results when all subagents complete their cycles
6. **Optional Quality Assurance**: If errors detected, run validation and mapping tools

## Subagent Count Determination (Asterisk Mode)

When user ends request with *, analyze complexity and determine subagent count:
- **Simple tasks** (2-3 subagents): Single-domain tasks, straightforward execution
- **Moderate complexity** (4-5 subagents): Multi-faceted tasks, some interdisciplinary aspects  
- **Complex tasks** (6-8 subagents): Highly complex, multiple perspectives needed

## Quality Assessment Criteria

Each Task subagent will assess task completion quality using these weighted criteria:

### Task Completion (40% weight)
- **Comprehensive coverage**: All aspects of the task addressed
- **Actionable outputs**: Clear, implementable deliverables
- **Adequate detail**: Sufficient depth without overwhelming complexity
- **Logical structure**: Clear organization and flow of information

### Accuracy & Relevance (30% weight)
- **Factual correctness**: Accurate information and data
- **Direct relevance**: Stays focused on the assigned task
- **No contradictions**: Internal consistency throughout
- **Current information**: Up-to-date best practices and standards

### Practical Value (30% weight)
- **Implementable solutions**: Real-world applicability
- **Clear next steps**: Specific action items and guidance
- **Resource considerations**: Realistic requirements and constraints
- **Success metrics**: Measurable outcomes and indicators

## Quality Scoring System

Each subagent calculates a weighted quality score (1-10):
- **9-10**: Exceptional quality, ready for immediate implementation
- **7-8**: Good quality, meets professional standards
- **5-6**: Adequate quality, needs minor improvements
- **3-4**: Below standards, requires significant improvement
- **1-2**: Poor quality, needs complete revision

**Quality Threshold**: 7.0 (subagents iterate until this threshold is met)

## Task Subagent Instructions

Each Task subagent will use the `scripts\gemini_client.py` script to spawn exactly 1 Gemini subagent for task completion:

1. **Generate Task Variants**: Create a specific approach for the task using these strategies:
   - **For 2-3 subagents**: Focus on different aspects (e.g., "strategy", "implementation", "evaluation")
   - **For 4-5 subagents**: Add perspectives (e.g., "planning", "execution", "risk management", "resource allocation")
   - **For 6-8 subagents**: Include specialized angles (e.g., "stakeholder analysis", "timeline planning", "budget considerations")
   
   **Examples for different task types**:
   
   **"Create marketing strategy" (3 subagents)**:
   - Task Subagent 1: "Strategic planning and market analysis for marketing strategy"
   - Task Subagent 2: "Implementation tactics and campaign execution for marketing strategy"
   - Task Subagent 3: "Performance metrics and evaluation framework for marketing strategy"
   
   **"Design mobile app architecture" (4 subagents)**:
   - Task Subagent 1: "Technical architecture and system design for mobile app"
   - Task Subagent 2: "User interface and user experience design for mobile app"
   - Task Subagent 3: "Security framework and data protection for mobile app"
   - Task Subagent 4: "Performance optimization and scalability for mobile app"
   
   **"Develop project management framework" (5 subagents)**:
   - Task Subagent 1: "Project planning and scheduling methodology"
   - Task Subagent 2: "Resource allocation and team management approach"
   - Task Subagent 3: "Risk assessment and mitigation strategies"
   - Task Subagent 4: "Communication and stakeholder engagement framework"
   - Task Subagent 5: "Quality assurance and success measurement criteria"

2. **Initial Task Execution**: Run `python scripts\gemini_client.py --delegate "specific task variant" --key N` (spaw