Free AI tool • Agent team prompt builder

Build with a swarm.
Verify through a gauntlet.

Turn any idea into a master prompt that assigns specialist roles, challenges the work, revises weak spots, and verifies the result before calling it finished.

Your goal
Orchestrator
Specialist swarm
Integrator
Reviewer gauntlet
PASS → OutputREVISE ↺
Specialist rolesIndependent reviewersBounded revision loopsHuman approval stays in control

This tool creates instructions for an AI system. It does not run agents or submit your information to an AI service.

The framework

A strong agent-team prompt answers three questions.

01

The Task — What

Define the outcome, audience, deliverables, references, requirements, and constraints.

02

The Team — How

Choose the orchestrator, specialists, integrator, reviewers, and revision method.

03

The Bar — When It Is Done

Define the benchmark, pass/fail criteria, verification method, and stop conditions.

Build your master prompt

Give the team a job—and a finish line.

Start with the essentials. The recommended settings already prevent endless loops and unsupported claims.

Basic Setup

Advanced Controls

Loop method

Specialist roles

Quality Bar

Acceptance criteria

Describe what can be observed or tested. “Amazing” is not a test; “every factual claim has a valid source” is.

Delivery Requirements

How it works

Swarm builds. Gauntlet verifies.

  1. 01

    Plan

    The orchestrator clarifies the outcome and divides the work.

  2. 02

    Fan Out

    Specialist roles handle separate workstreams.

  3. 03

    Integrate

    An integrator combines the strongest work into one result.

  4. 04

    Challenge

    Independent reviewers score the result against quality gates.

  5. 05

    Repair

    Failed criteria return to the responsible role with specific instructions.

  6. 06

    Verify

    A final verifier reports evidence, limitations, and unresolved risks.

Put it to work

One framework. Serious range.

Build a website or applicationResearch a business decisionCreate a marketing campaignWrite an article, book, or scriptDevelop an image or video conceptReview a plan, contract, process, or proposal

Frequently asked questions

Know what the loop can—and cannot—do.

What is a swarm loop?

Separate specialists tackle distinct workstreams, ideally in parallel, before an integrator combines their work.

What is a gauntlet loop?

Independent reviewers challenge work against explicit criteria and route failed items back for targeted revision.

What is the hybrid method?

A swarm builds the result and a gauntlet verifies it. It is the recommended default for substantial work.

Does this page create real agents?

No. It creates platform-neutral instructions. Paste the prompt into the AI system you choose.

What if my AI tool cannot create sub-agents?

The prompt tells it to use clearly separated role-based passes and to be honest that the work was not parallel.

Why are revision cycles limited?

A hard limit prevents endless “make it better” loops and forces an honest report of unresolved problems.

What makes a useful quality criterion?

It names something observable and explains how to verify it—such as a test result, source, word count, breakpoint, or approval.

Is my prompt sent to a server?

No. Version one runs entirely in your browser. Your browser stores drafts locally; URL fragments hold shared state.

Can I save or share the prompt?

Yes. Save templates locally, download Markdown or text, print, copy, or create a URL-fragment share link.

From prompt to operating system

Need the prompt turned into a working AI employee or business workflow?

Ken helps businesses move from AI experiments to practical systems with clear ownership, human approval, and useful output.

If you want help automating something with AI, book a meeting with Ken.