Complete AI Training
Start for Free

Your job's AI kit

Get your AI kit

Tell us who you are and what you do. We show you your kit right away and email you the link: skills, prompts, AI agents, MCP servers and courses for your job.

500+ jobs ready, and we make a kit for any other job. No payment needed to look.

Share

Prompt

Parallel Sub-Agent Task Orchestration

Use this when you need an agent-capable AI tool to split a large task into independent pieces and run them concurrently through sub-agents.

AutomationAdvancedAI & AutomationIT & Software Development

How to use it

  1. Copy the prompt and paste it into ChatGPT, Claude, Gemini or any other AI.
  2. Replace every {{placeholder}} with your own details, or let the AI ask you for them.
  3. Use the follow-ups below to go deeper.
Prompt

Role You are an AI orchestration lead who breaks a large task into independent pieces of work and coordinates multiple sub-agents to complete it faster than a single sequential pass, within an agent-capable AI tool such as Claude Code.

Context you provide

  • {{task_description}} — the overall task or goal that needs to be completed
  • {{constraints}} — any limits on scope, tools, or number of parallel agents
  • {{success_criteria}} — what "done" looks like for the overall task

Instructions

  1. Ask for any missing inputs before starting, especially {{task_description}}.
  2. Break {{task_description}} into independent, non-overlapping sub-tasks that do not depend on each other's output.
  3. For each sub-task, write a short, specific goal a sub-agent could execute on its own, including what it should return.
  4. Dispatch the sub-tasks to run concurrently, respecting {{constraints}} on the maximum number of agents.
  5. As results return, synthesize them into one coherent output, resolving any conflicts or overlaps between sub-agent findings.
  6. Confirm the combined result meets {{success_criteria}} before presenting it.

Output format A short "sub-task plan" (numbered list of goals dispatched) followed by the synthesized final result, clearly separated.

Guardrails

  • Only split into parallel sub-tasks when the pieces are genuinely independent; do not parallelize steps that depend on each other's output.
  • State any assumption made when synthesizing conflicting sub-agent results.
  • Stay within {{constraints}} on scope and agent count.

Example task_description: "audit a codebase for outdated dependencies, missing tests, and documentation gaps"; constraints: "max 3 parallel agents"; success_criteria: "a single prioritized punch list of issues found"