OpenClaw Skills + Base

Hi everyone, Si here. So in the last video, we managed to get our agent up and running. So we set up our own server so our agent can run in its own environment. Then we were able to prompt the agent for a wallet and also do some very basic transactions. In today's video, we're going to be talking about skills, how you can install them, how you can make them, and I'll be demoing a new skill that I've made myself.

Just a word of advice. This video is not intended for financial advice. I am just showing everybody how you are able to build and extend your your agents. So, NFA 100%. So, what is a skill?

So, we were able to prompt quite far, but you don't necessarily want to leave everything to chance. For example, if I wanted my agent to do something the same way every day, I couldn't guarantee that just by a prompt. I'd get a lot closer by using a skill. Think of skills as recipe for your agents, these predetermined railways for them to do a task. So, here I am.

I'm talking to my agent and I asked for it to list its skills. So we have weather the skill creator blue bubbles and then you can see as the agent says right here that base trading skill is a custom skill that they built well we built it together and and then the others just come with open core. [snorts] What what the agent means is that when you first start your open core agent, there are a bunch of skills that come with it by default, which I think is really neat. But [snorts] then, of course, you can extend it. Now, if you're like me, you've probably been building a few skills.

And now that you have a secure environment for your agent, you want your agent to have that skill. So, how can I get this skill that I have on my desktop to my agent? I will show you in 1 minute. You can easily send file files from your desktop to your remote server by sshing into your remote server and then determining where you want the skill to go. That's what I did with my skill here, the base trading skill that my agent and I got to fix a bit because we tested it out.

Now in our skill MD file, you will see that you have the metadata at the top and then the rest is just instructions for the different assets and references and scripts that we have as part of our skill. [snorts] The name and description is so incredibly important because this is how the agent knows to call this particular skill.

[snorts]

It will look for terms like trade tokens on base chain, handle swaps or DCA positions and limit orders. All of these are triggers for the agent to call up the skill. So your open core agent can list out it out its skills like it has here. But the other really cool thing that will happen and that we just touched on is that the skills load on demand. So that means that your agent can have access to hundreds of capabilities, but it will never but it doesn't have to bloat the context window because it will only call that skill when it hears the trigger words.

Ask why did I focus on building a skill that DCAs or uh limit orders? So the first one is that with dollar cost averaging it spreads a purchase over time. So, it's a really good it's a really good skill to to see that you can sort of set certain certain time frames for your agent to run the script. And a limit buy will set a target price and the order will only be filled by the market when it hits that number. So, both of these skills have both of these tasks have one thing in common, the crunch jobs or so it means that I'm able to give a time frame.

The agent can then while I'm sleeping run this run these checks have this work done for me while I am sleeping which I think is really cool. So let's see the skill in action. So I'm just asking hey agent let's use the space trading skill. The agent gives me some prompts that I should use. For the purpose of this tutorial, I'm going to DCA.

So, I've asked the agent to buy me um $3 worth of ETH on base for the next hour. It correctly told me that right now the skill really only supports um daily or weekly frequencies, not hours. And that makes sense because that's that's not what you want to do. But for the purpose of this demo, it works.

[snorts]

And so we worked out, we both negotiated and we worked out to buy $125 and it ran the script and bought me some ETH. So let's check that out. So here you can see it did the swap correctly. Everything's looking good. So that is like that is one example of this of this scale.

So I'm going to ask our agent to buy us Ethon base if the price drops below the current price. Great. So, the limit order um [snorts] needs to run periodically to check the prices and it asks me if I want to set up the crunch job to monitor. So, this is all fun, but what if you want to make your own skill? Well, let's see what the trending tokens on base are.

So, I asked it what the current trending tokens are in base and to list them by volume and list the top five.

[snorts]

That's amazing. What if we wanted to make this a skill? This is where the fun comes in. And there you have it. So the agent was able to call um was able to add a trending script under scripts.

And if I want to know now what the trending tokens are, it's just part of trending skill. So, one thing I needed to correct is that the agent did actually use the wrong endpoint to get that data. And then all I was able what I was able to do is just to ask it to to check its data. I was able to to give back the correct information. So, if you have that situation ever happen to you, please always check.

It's it's just a good lesson to have. Your agent can make not things up but it can use different endpoints that it shouldn't. So always validate what comes out of your agent. Do not just take it as gospel and then it is able to go and update the skills.mmd so it will be correct.

If you want to learn more about creating more in-depth skills, I highly recommend using the Claude code documentation on skills. The skill creator that uh Claude that Open Claw uses is actually from Anthropic. So, you can learn all about how it works, how to set it up, what to look for in a skill. Everything is there. And even better, it's all open source.

I can't I can't wait to see you guys in the next video where our agent's getting kind of lonely. So, I wonder what it'd be like if he had some friends. Thank you so much. Bye, guys.