📹 Clone My Playbook: Replicate My AI Agents in Your Own Environment
Jules Boiteux
4 months ago16 views
Transcript
00:00
All right, so now that we have set up cloud code and started creating some sub agents on our local environments, I'm going to show you how you are going to actually like steal my own agents to actually add them to your own setup.
00:19
So as we saw we now have like the agents, the corresponding custom commands and also the rag, contents that the agents are going to learn and get inspiration from.
00:32
now obviously it takes some time to add up all those agents but the good thing is that as I said I share with you the content of all my sub agents.
00:41
So let's actually copy one and let's actually copy the refine LinkedIn article that I showed you in the first video.
00:49
Okay, so the code agent.
00:52
So here is the custom commands.
00:55
So it's called Refining Article.
00:57
It's in the LinkedIn folder and as you can see it links to that specific cloud code agent that I have and for which I have like the corresponding like instructions.
01:09
Okay, so here we go.
01:10
What we're going to do is as usual, we are going to start with setting up the agent.
01:16
We're going to create the agent here.
01:18
and this time instead of generating it with cloud, we are going to use manual configuration.
01:24
So what's the unique identifier for the agent?
01:26
Well, this is this name here.
01:29
Right here.
01:31
the question is enter the system prompt for your agent.
01:34
So that's basically like the instructions and the instructions, this is actually the core of the content here, that you can directly copy Paste.
01:43
Right now claude code is asking when should claude use this agent?
01:49
Well that's exactly the goal of this description here, right?
01:56
And then once it is done you can hit continue here, use sonnet again, choose the color you want and then save it.
02:04
Right?
02:05
And as you can see we now have the agents that has been added here.
02:11
as you can see obviously there's specific instructions related to what is the folder to use.
02:20
And here the instruction is, okay, you should fetch existing potential existing content inside this specific inside this specific folder.
02:34
And then you should write up also content in, inside this specific folder as well.
02:39
So there are actually two changes to make.
02:41
The first one is first let's actually add up this new folder, okay that we're going to call LinkedIn articles.
02:52
And the thing is that in the wake of the demo I used another name for the folder here.
02:57
So I'm actually going to Rename this one to Obsidian so that you know, the folder that I use myself does have the exact name.
03:08
That also means that I need to also update the folder that is used inside the if email refiner, you know, agents.
03:17
But that's something that I'm going to give as add the instructions later.
03:21
Anyways.
03:22
What's happening is that I have this new agent here.
03:28
So what I'm going to do is that I'm going to actually add up those more specific instructions and say I have renamed the folder Obsidian.
03:39
So make sure that the instructions inside refiner, sorry email inside email refiner.
03:50
And LinkedIn article does take this renaming into account.
04:01
also make sure for LinkedIn refiner that name of the folder that is used to add up those generated articles is LinkedIn.
04:23
Okay, I don't have it is LinkedIn underscore articles.
04:27
Okay, so yeah, this is a kind of refinement that you'll probably need to do just to adapt, you know, the agents that have for you to your proper structure.
04:37
And while it's actually working, let's actually add manually the command that is associated to the LinkedIn article refiner, agent.
04:48
So it's very easy if you jump to the folder called Comments, you can just add a new file and then you can call it with the naming that I use and the naming that I use for this specific agent is actually refine LinkedIn article.
05:04
Okay, so let's actually proceed with this and let's actually recreate the file.
05:11
Okay.
05:12
It's called Refine LinkedIn Article MD.
05:15
And I'm just going to very simply.
05:17
Oops, sorry, I'm going to very simply like copy the contents.
05:22
Right.
05:23
So as you can see, it's performing the EDI edit that I wanted that is adapting the name of we are referring to to this, you know, very specific name.
05:36
And yeah, now that we have done that, we now have the custom command that has been created.
05:42
We now have the agents, that is referred to once the custom command is called.
05:48
And we of course have the folders that are going to act as you know, sources of additional information for the agents to adapt the content here.
06:00
So now we just need to actually hit Control C on Mac twice just to close out the cloth commands, restart it again and then what's going to happen is that if I hit the refine linear active comments, I can see that I have those two arguments that are put forward here.
06:23
And I'm actually going to add up the argument here which is basically what I added in a former video.
06:37
I'm going to provide some more instruction, make it sharp and with a great hook.
06:44
So obviously that's a very shallow additional instructions but that's for the sake of the demonstration.
06:51
And yeah as you can see cloud code is going to on refining the LinkedIn article Right.
06:58
And I see now that the article has been generated so it is now available in the corresponding directory.
07:05
So yeah, that's the process to actually replicate the agent stats and custom that I have shared with you.
07:15
So yeah, next step is for you to just mimic this setup and just steal the content from my agents.
07:23
And what I advise you to do is obviously to update those agents, to improve them as you use them and pinpoint what should be improved and also make sure to add more of your own custom commands or sub agents whenever you realize that you are performing the same kind of tasks or deem that cloud codes help, can really improve either time spent on the task or the output or even both.
08:01
So yeah, thanks a lot for following up on this content and yeah, don't hesitate to provide me with any feedback you have.
08:09
always happy to keep improving my setup with the community.