YouTube Deep Summary

All right, so Claude Code just keeps delivering. So Sid put a post out. He's one of the engineers at Anthropic. And when I saw him and followed him earlier today, it was around 400 followers. You can see that this post has already got over 390,000 views. And Cloud Code has just released the coolest feature yet. It's called agent mode. So now you can have a contextbound agent directly in your Cloud Code terminal. So you can give them very specific specialized tasks. As you see here, it's showing a demonstration of a back-end architect, a front-end developer, and it just goes on and on. So, I haven't seen any posts of anyone actually putting some of their system prompts out. So, we're going to do this live here. So, first thing we can do here, if you go to the Anthropic documentation, you'll see that it walks through how to create a new agent, where it's located, whether a project-based agent or one for your user that is global for your entire system. So, what we're going to do here is why not use Claude to create the Claude agents. So, I went over and I created a system prompt here that says you're an expert in writing highly effective LM system prompts. So, I'm not going to read through all of this, but essentially I'm giving the documentation that says, "Hey, Claude, go look at your own sub agent documentation and then I'd like you to reference Superclaude." And what Supercloud is is this is super great. Like before we had agent mode, we were using Superclaude here to have these different personas. So the personas of an architect, the personas of a security engineer, personas of a front-end engineer, we already have some of this this information baked in and supercloud also comes with key principles. So these are software development core principles for coding standards that you have here and rules of engagement to walk through what are some of the actionable rules that you would have as part of these. So, as part of this prompt that we're writing here in Claude, it's going to go ahead here and build out some of the different rules and ways that we can go ahead and build out some of these sub aents directly in our Cloud Code interface. So, we're going to pause here and come back as soon as it's finished. All right, so we're finished now and we have nine sub aents that we've built. So, I've prompted Claude to go ahead and create markdown files. Each of these stored as individual artifacts that we can use. So, as you can see, we have all nine different personas that are available to us. So, we're only going to use one here. Let's look at the architect sub agent. And you can see that we can go ahead here and copy this directly into our clipboard. And it goes into all the key information that I was able to discover from looking at Superclaude in the relevant documentation. So, what we'll do next is we're going to go ahead here over into Claude Code. So to create a new agent, you're going to hit the / agent command and then we're going to click create new agent. We're going to make this a personal agent so we can use it in all of our projects that are available to us. So we'll click on that and we're going to follow the documentation which is the recommended to generate with Claude. Once we go through, we're going to go ahead here and we're going to paste in our system prompt for our architect sub agent. And we're going to hit enter. And what this is going to do is it's going to go ahead and generate our agent configuration. So we'll come back here as soon as it's finished. All right. So now the configuration has been created and we're at the step where we can give it tools. So for this one we're going to give it all tools just for the example. Maybe for an architect agent we want to just have it as read only. So you're only using sequential thinking and context 7 for information gathering for the knowledge bases. But for this we'll go ahead and hit all tools here. So, we'll go ahead and make sure that that's selected and we'll click on continue. And then you can change your background. So, for this, let's just say that we're going to use the system architect is going to be a blue agent. All right. And then we're going to get to our final steps. In which case, we're going to go ahead here and it has a name of system agent and it's been created. So, now we have the ability to use that in our codebase. So, what we're going to go ahead and do is I'm going to start going through and set up the rest of these other eight agents and I'll do another post here in the future where we talk about some of the practical use cases that we can use these sub agents for. And what is super cool when we were actually looking through some of the post here from what Sid posted on his X feed, it does look like it's going to use the base model that you have. So, you're going to have access to both Opus and Sonnet as long as you have that as your default model. it's not going to be utilizing the haiku models for execution. So, I hope this is super helpful and I'm I'm looking to see in the comments of any other use cases that you guys are finding for different sub agents.