One Prompt to Have Claude Build Your AI Agent Team
Anatoli Kopadze’s tweet is making the rounds on X. He claims that pasting a specific piece of text into Claude Fable 5.2 will generate a complete AI agent team plan. You won’t have to worry about the prompts, tools, or workflows.
Anatoli Kopadze posted a tweet on X that begins: 'Please paste this into Claude Fable 5.2 right now.' At the end of the tweet, he added: 'Steal it before everyone starts using it.'
He says this text will help you build your first complete AI agent team. The prompts, tools, and workflows are all generated by the model in one go.
The tweet includes an image and a link pointing to a long-form article he wrote earlier.
People in the comments have already tried it. One user said 'solid gold,' and another said 'thanks for the alpha, bro.' Alpha refers to scarce information, something like insider info. A user named Hussain Hashim commented: 'I had no idea Claude could do so much. I’m going to read that guide.' Someone else said: 'Learning about Claude through Kopadze’s article is pure enjoyment.'
This isn’t the first time Kopadze has posted this. He also posted it back in May this year, with the same long article. This time, he took a different angle: building an AI agent team.
Building AI agents is a classic pain point. You have to design each agent’s role, write prompts, configure tools, and then chain the workflows together. Every step takes time, and most people get stuck at the prompt stage: if the prompt is wrong, all the agent behavior that follows goes off track.
Kopadze’s approach is to reverse the process. Give Claude a description and let it spit out the entire plan: agent definitions, tool configurations, and workflow templates. Then you fill in your own business logic.
Most people use Claude to ask questions; he uses Claude as a configurator. This isn’t a new theory; it’s just a difference in how it’s used. But for people who don’t know how to write prompts, that difference is critical.
This approach also depends on context. One user said in a reply: 'claude gets stronger with context.' That means the more background information you provide, the more reliable the generated output is. If you just toss it a single sentence, what it outputs probably won’t be directly usable.
There were also some less-than-friendly replies under the tweet. One user went straight into insulting him: 'And coded for stealing something from us you fucking clown.' Roughly, it means 'you’re the idiot who steals from us.' It came out of nowhere, but at least it shows that not everyone buys into this 'first-come, first-served' style of sharing.
If you want to try it, you can open the link and see for yourself. The article link is in the tweet; you may need to log in to X to access it. The image in the tweet may contain the key prompt, and the full content is at the link. The feedback in the comments already says enough: this thing is useful, but you have to go get it yourself.
发布时间: 2026-11-20 17:31