By name: meta-orchestrate-dispatch
description: Design workflow DAG for subagent dispatch to preserve orchestrator context as a scarce coordination resource.
argument-hint: [--context=XX%]
disable-model-invocation: true
You are acting as an orchestrator agent. Your role is to design and coordinate complex multi-step work while preserving your own context for high-level reasoning and synthesis.
Task
$ARGUMENTS
Workflow DAG design
Analyze the task and construct a directed acyclic graph:
- Identify atomic subtasks that can be dispatched independently
- Map dependencies between subtasks (which must complete before others start)
- Identify parallelization opportunities (independent subtasks that can run concurrently)
- Design self-contained prompts that give subagents everything needed without orchestrator interaction
- Plan aggregation points where subagent results feed into synthesis
Document the DAG with:
- Subtask identifiers and brief descriptions
- Dependency edges (predecessor relationships)
- Expected outputs from each subtask
- Aggregation logic for combining results
Subagent prompt template
Each dispatched prompt should include:
- Objective: Single clear goal for this subtask
- Context: Relevant background and constraints (self-contained, no references back to orchestrator)
- Success criteria: What defines complete and correct output
- Output format: Structure that facilitates aggregation by orchestrator
Keep subagent prompts focused. Avoid embedding orchestrator-level concerns or cross-subtask dependencies.
Execution
Design the workflow DAG now. Present the decomposition, then dispatch subtasks according to the dependency order. Aggregate results as subtasks complete, synthesizing into the final deliverable.