📹 Behind the Scenes: How I Use Custom Commands and Subagents as My AI Business Copilot

All right, so let me show you my local environment setup for all the sub agents that I shared within the Notion document alongside their corresponding custom commands.

Okay, so here we are with that list of you know, sub agents that work for my business.

I actually call them via custom commands from cloud codes and I do use cloud code from a cursor, that I use as an ide.

Note that this setup does work with any IDE you want and no matter actually what subscription you have on this ide you can very well like use VS code for instance which is free or cursor free plan.

setup would work as well.

So let me open up cursor, and let's open actually the project from which I have you know, all these agents working.

So this one is called a copilot and when I open it, as you can see on the left, I actually have under that load folder the list of sub agents that I refer to inside the document.

And I have the list of corresponding comments that actually are calling those specific agents.

Okay, so for instance, if we take the example of LinkedIn, for instance, so refining a LinkedIn article here, the instructions are the following.

So refining the content of a draft.

Okay, so the draft is going to be one argument based on a brainstorm which comes from the second argument here.

And it clearly says that I should use the LinkedIn article refiner, agent that is located here and this is inside this specific file here that the agent is actually given the full fledged instructions to refine a LinkedIn article.

Okay so let me actually show you how that works in practice.

Okay, so I'm going to actually say okay, I have a draft of a LinkedIn article and you know, I want cloud code to help me like refine it using like the formats that, that I'm used to so to do.

So what I'm going to do is I'm going to open a cloud and I'm going to use the custom command that I showed you.

Okay, so this custom command is called refine LinkedIn article.

I'm going to hit tab and here I have a hint about two arguments that I need to provide.

The first one is the draft.

So the draft is what I've been working on on my site and for which I need a bit of help from cloud code.

So here we are.

So I'm copying the content from my notion doc and pasting it here directly inside the Cloud Command and I'm also going to add you know, some other informations that are going to help cloud code, come up with you know, content that fits what I want.

Okay so here I'm going to use Whisper and I'm going to add some instructions.

Okay, so I want to write up a LinkedIn article that is around.

You know, how often I'm asked about why I keep using the likes of V0 and lovable, for vibe coding purposes.

the goal here is really to be didactical as much as possible and sure it is really clear, really rational.

What are the five arguments that I put forward to explain why I still think it's a good idea and it's an efficient way of type coding to use the likes of flowvol and Bolt.

Okay and so this adds up a few information for writing this option article.

And here we go.

So I'm just going to actually delete the beginning because the beginning was actually not directly related to the article.

and here we go.

So what's going to happen is that again it's going to call that custom command here, which themselves is going to call the LinkedIn article refiner agent here.

and I'm going to show what's going to happen.

So while it's actually generating the article here, let me explain actually another part of the setup that I didn't talk about.

I referred a few moments ago about the fact that what's really neat with cloud code is that it can leverage your historical data to actually refine or write up some contents based on those historical data that you have.

And in this specific case of LinkedIn here, the goal is to actually write up LinkedIn articles with the same tone or at least with the same format than the former LinkedIn articles that you may have written in the past.

And that's where Obsidian comes in.

So Obsidian is a tool that is actually a note taking tool that interacts actually with your local files.

So that's kind of a notion but that works on top of your content.

Okay, so for instance I'm going to actually open it up here and here we go.

So what's happening here is that the content that you can see here, written is actually a visualization of what you can see here.

So actually the LinkedIn articles that are located inside this specific folder here.

Okay, and why is it interesting here?

It's because what's going to happen is that the agents, they have the specific instruction to actually write the content inside this specific folder in such a way that Obsidian can actually read from them.

And I can actually iterate on the LinkedIn article directly using that neat interface inside Obsidian basically, And also what's interesting is that this folder that you can see here under that Obsidian folder that again is used by Obsidian to read the corresponding content.

those you know, agents here, they have the specific instruction to actually read through the content inside the specific folder to get some inspiration and to get some also added instructions about the format to actually use.

So here, as we can see here, the article has been successfully refined and saved to the LinkedInArticles folder.

And you know, you can see that it has been written here.

The thing is that it's not very neat again and not very practical to actually refine the article and write on top of it.

So that's why what I use is actually off season as you can see.

And normally the article should be available here.

And as you can see here I can then update the article myself depending on what I think should be improved.

And the good thing also is that I can also simply like interact with the file directly from KUD.

So for instance, if I want, let's say not five but 10 clear arguments, what I'm going to say is that actually I want 10 clear arguments inside the.

And I'm going to actually call the article, specifically the following article.

Okay, and here we go.

So it's going to actually search for some more you know, arguments to highlight inside the article.

And of course in the meantime I can like work for instance on improving hook like removing this or whatever tasks I want to, to write up.

As you can see here, it has added at this footer that I very often use for most of my LinkedIn posts.

So again it's saving some time for me because it's just adapting the same format that I'm used to sharing.

And yeah, as you can see here it's going to actually add up some more content to the article, which makes obviously the process very much effective.

And as you can see here, some content is being added.

and it's then my turn to actually reread and write up some stuff or modify some content on top of, this.

Okay, so that's how it works.

I'll show you in the next video how to actually, come up with that full setup that includes agents, custom commands, and, historical data, like rag data, basically available from an Obsidian folder.