Planner–Worker–Reviewer Teams

Any modelAutomation From a course

The most robust agent architecture is also the simplest: a Planner decomposes the goal into tasks, Workers execute one task each with narrow context, and a Reviewer checks results against acceptance criteria before anything ships.

The prompt

PLANNER: Decompose this goal into 3-5 independent tasks with acceptance criteria each. Goal: [GOAL].
WORKER (per task): Complete exactly this task, nothing more: [TASK + CRITERIA].
REVIEWER: Check this result against the criteria. Verdict: PASS or FAIL + one-line reason. Result: [RESULT]

What to replace

Swap these placeholders for your own details before running the prompt:

[GOAL]your own value
[TASK + CRITERIA]your own value
[RESULT]your own value

Pro tip: The Reviewer must be a separate call with fresh context — models grade their own work far too generously.

How to use this prompt

Copy the prompt using the button above.
Replace [GOAL], [TASK + CRITERIA], [RESULT] with your own details — the more specific you are, the better the output.
Paste it into Any model and run it.
If the answer feels generic, add constraints: audience, length, tone, and what to avoid. That single change fixes most weak output.

Learn the technique

Planner–Worker–Reviewer Teams

Module 3 — Advanced: Multi-Agent Patterns · Loop & Agentic Engineering

Related prompts

ChatGPT

Critique-and-Revise Loops — ChatGPT

A loop is simply feeding a model's own output back to it (or to a second prompt) for review and improvement, repeated until a quality bar is met.

Claude

Critique-and-Revise Loops — Claude

A loop is simply feeding a model's own output back to it (or to a second prompt) for review and improvement, repeated until a quality bar is met.

ChatGPT / Agent tools

The ReAct Pattern (Reason + Act)

ReAct loops interleave reasoning ("what should I do next?") with actions (calling a tool, searching, running code), then feed the result back in before reasoning again.

Claude

Giving Loops Memory

For loops that run over many steps (research agents, multi-turn assistants), summarize prior steps into a short running memory instead of replaying the full history every time — this keeps context small and cheap..

Any model

Loop Budgets and Failure Handling

Production loops need budgets (max iterations, max cost, max time) and explicit failure paths: what happens when the Reviewer fails a result three times? Options: escalate to a human, fall back to a simpler method, or return a partial result flagged as unverified.

Any model

Project: Draft–Critique–Revise Content Machine

Deliverable: a working 3-role loop (writer, critic, editor) that produces a publishable piece of content on any topic you give it. Steps: 1.