100 Claude Agents From 1 Prompt: The GTM Content Production Playbook

Most GTM engineers use the word 'agents' without understanding that Claude has three distinct architectures. Using the wrong one for the wrong task wastes time and tokens. This page explains the three types and when to use each.

Sub-agents: the original architecture

Claude holds the plan in its head and launches a few helpers

Agent teams: structured roles, direct communication

You define the team. Agents talk to each other. Claude orchestrates.

Dynamic workflows: Claude writes the plan, launches up to 1,000 agents

You describe the objective. Claude writes the orchestration script. The script manages everything.

When to use each

Match architecture to task size and structure

The trigger for dynamic workflows is the word 'workflow' in your prompt. Three methods to activate it. One additional setting (Ultra Code) that lets Claude decide when to use workflows automatically.

Method 1: The trigger word

Include the word 'workflow' anywhere in your prompt

Method 2: The effort slider

Type /effort in terminal to access all effort levels including Ultra Code

Method 3: /workflow command

Type /workflow in Claude Code terminal to see running workflows

How to write the prompt

Describe the objective and specify the phase structure if you want control over how work is divided

A live demo from the video: 100 agents deployed to produce a complete YouTube content package. Four phases, each running in parallel. The structure maps directly onto any content calendar, campaign brief, or ICP-targeted content production task.

The exact prompt

Copy-paste this. Adapt the topic and your channel name.

The objective and prompt structure

One prompt. Four phases. 100 agents. 6 minutes.

The judge phase: why it matters for GTM work

10 agents with different perspectives vote on the best outputs. Only the strongest survive.

Adapting the demo for GTM content operations

The four-phase structure maps directly onto any content batch operation

The actual GTM prompts — copy and paste these

Four ready-to-run prompts for GTM content operations

The most counterintuitive thing about dynamic workflows: 100 agents each consuming tokens does not fill your context window. Your main session used 7% context and 75,000 tokens despite 100 agents each using approximately 20,000 tokens. This is not an accident of the demo. It is a fundamental architectural feature.

How this is possible

The orchestration script runs outside your session. Only the final result returns.

What this means for long-running GTM operations

You can run large batches without context degradation limiting quality

The cost reality

Agents outside your session still burn tokens. You just don't see them in your context.

AI Driven, Personalized Outreach on LinkedIn: Prosp 💜

↳ Start a 14 Day Free Trial 💜 ↳ Watch Full Tutorial ▶️