task-orchestra（任务乐团）

Task Orchestra

Coordinate multiple agents and tasks for complex workflows.

When to Use

• Multi-step operations requiring coordination
• Parallel execution of independent tasks
• Complex workflows with dependencies
• Orchestrating subagents for large projects

Core Capabilities

1. Task Coordination

• Break down complex tasks into manageable steps
• Manage dependencies between tasks
• Coordinate parallel execution
• Handle task sequencing and scheduling

2. Agent Orchestration

• Spawn and manage multiple subagents
• Route tasks to appropriate agents
• Monitor progress and handle failures
• Aggregate results from multiple sources

3. Workflow Management

• Define workflow patterns and templates
• Implement error handling and recovery
• Manage state and progress tracking
• Coordinate handoffs between agents

4. Dependency Resolution

• Analyze task dependencies
• Create execution order
• Handle conditional execution
• Manage resource conflicts

Orchestration Patterns

1. Sequential Execution

Task A → Task B → Task C

2. Parallel Execution

Task A, Task B, Task C → Aggregate

3. Pipeline Processing

Input → Task A → Task B → Task C → Output

4. Supervisor Pattern

Coordinator → Multiple Subagents → Results

5. Event-Driven Processing

Event → Trigger → Response → Next Event

Quick Actions

• orchestrate [workflow] - Execute complex workflow
• parallel [tasks] - Run tasks in parallel
• pipeline [steps] - Chain tasks in sequence
• supervise [agents] - Manage multiple agents
• dependencies [tasks] - Analyze and resolve dependencies

Usage Examples

"Orchestrate a complete research project with multiple agents"
"Run these tasks in parallel and combine results"
"Create a pipeline for content creation from research to publication"
"Supervise a team of agents working on different aspects"
"Analyze dependencies and create execution order"

Workflow Templates

Research Project

1. Research Topic → Research Agent
2. Data Collection → Data Agent
3. Analysis → Analysis Agent
4. Report Generation → Writing Agent
5. Review → QA Agent

Content Creation

1. Topic Research → Research Agent
2. Outline Creation → Writing Agent
3. Draft Writing → Writing Agent
4. Editing → Editing Agent
5. Publication → Publishing Agent

Software Development

1. Requirements → Analysis Agent
2. Design → Design Agent
3. Implementation → Coding Agent
4. Testing → QA Agent
5. Deployment → Deployment Agent

Agent Management

Spawning Agents

sessions_spawn({ task: "specific task", label: "agent-name", mode: "run" })

Monitoring Progress

subagents list

Handling Failures

subagents kill [agent-id]
subagents steer [agent-id] "new instructions"

Dependency Resolution

Types of Dependencies

• Data Dependencies: Task B needs output from Task A
• Resource Dependencies: Tasks sharing same resources
• Order Dependencies: Tasks must run in specific order
• Conditional Dependencies: Task runs only if condition met

Resolution Process

1. Identify all dependencies
2. Create dependency graph
3. Find topological sort
4. Execute in dependency order
5. Handle conflicts and cycles

Error Handling

Common Failure Scenarios

• Agent Failure: Subagent crashes or times out
• Dependency Failure: Required task fails
• Resource Conflict: Multiple agents need same resource
• Network Issues: API calls fail or timeout

Recovery Strategies

• Retry: Attempt failed task again
• Alternative: Use different approach or agent
• Skip: Continue without failed task
• Rollback: Undo previous steps

State Management

Progress Tracking

• Track completed tasks
• Monitor current execution
• Record task results
• Maintain workflow state

Checkpointing

• Save progress at key points
• Enable restart from checkpoints
• Maintain consistency across failures

Communication Patterns

Parent → Child

/sessions_send [agent-id] "instructions"

Child → Parent

Auto-announce results
Reply with findings
Report errors and status

Agent → Agent

Share data through files
Coordinate via shared state
Trigger other agents

Performance Optimization

Parallel Execution

• Identify independent tasks
• Run in parallel when possible
• Aggregate results efficiently

Resource Management

• Monitor agent resource usage
• Balance load across agents
• Avoid resource conflicts

Efficiency Metrics

• Task completion time
• Resource utilization
• Error rates
• Success rates

Safety Considerations

Agent Limits

• Max 10 concurrent subagents
• Max 2 levels of nesting
• 10-minute timeout per agent
• Automatic cleanup

Data Integrity

• Validate task inputs/outputs
• Maintain consistency
• Handle partial failures
• Ensure atomic operations

Advanced Patterns

1. Hierarchical Orchestration

Main Coordinator → Team Coordinators → Individual Agents

2. Dynamic Work Allocation

Assign tasks based on agent capabilities
Reassign if agent fails
Balance load dynamically

3. Event-Driven Workflows

Event → Trigger → Agent → Result → Next Event

4. Adaptive Planning

Plan → Execute → Monitor → Adjust → Repeat

Integration with Other Skills

Self-Evolution

• Use for complex self-improvement tasks
• Coordinate multiple evolution agents
• Manage long-term capability building

Analysis Skills

• Orchestrate research projects
• Coordinate data analysis
• Manage multi-step investigations

Content Creation

• Coordinate content production pipelines
• Manage multi-agent content creation
• Orchestrate publication workflows

Quick Reference

Common Commands

# List running agents
subagents list

# Kill failed agent
subagents kill [id]

# Send instructions
sessions_send [agent-id] "message"

# Spawn new agent
sessions_spawn({ task: "task", label: "name", mode: "run" })

Workflow Examples

# Research project
orchestrate "research-project" with agents: research, analysis, writing

# Content pipeline
pipeline "content-creation" with steps: research, outline, draft, edit, publish

# Software development
supervise "dev-team" with agents: analysis, design, coding, testing, deployment

Best Practices

1. Start Simple: Begin with sequential execution
2. Add Parallelism: Identify independent tasks
3. Handle Failures: Implement robust error handling
4. Monitor Progress: Track execution and results
5. Optimize Performance: Balance load and resources

Success Metrics

• Task completion rate
• Execution time efficiency
• Resource utilization
• Error recovery effectiveness
• Overall workflow success

Remember: Good orchestration makes complex tasks manageable and reliable.