From prompt to Python: building Launchpad custom functions with Agent Skills

Recently, I was working on a project where I needed to be able to load financial filing data for US companies in my workflow to apply generative AI and decisioning to it. 

I looked around at the options, including scraping stock websites, parsing HTML, and none of those things sounded appealing to me. I want to spend my valuable time building my SaaS product, not writing a bunch of code to parse data. 

I finally stumbled upon a very useful python library: yfinance 

But how could I use this timesaving python library in my Launchpad application? 

Turns out this was as easy as installing the new launchpad-agent-public-skills into my VSCode workspace with this command: 

 

Once they were installed, I just gave github copilot a simple request:  

 

Within a few seconds, I had a new python module and instructions on how to create a Function rule in my Launchpad application to take advantage of it: 

 

 

What did my coding agent generate based on the launchpad-custom-function skill? Everything I need:

1. A stock_lookup.py module with a lookup method 

2. A python requirements file describing the other python libraries needed 

3. A build script (in powershell because I’m on windows) for installing the dependencies and generating the zip file needed to create a Function rule in Launchpad. 

 

After running it locally just to double check, I was good to go! But first, I wanted to make sure I understood the next steps, so I asked my coding agent: 

 

The response used the skill to guide me through the next steps, including the fact that I really wanted to have two function rules, one for each method, by telling me to: 

1. Create a new Function rule 

2. Choose the Python 3.12 runtime 

3. Upload the generated zip file from my workspace 

4. Add one input parameter: symbol 

5. Set the output to text for testing 

6. Use the Run rule tool to try it out 

 

So, I created the Function rule and tried it out on Launchpad. It works!  

 

Next steps 

Now that I’ve built the python code I need, by taking advantage of open-source python libraries to access stock data, I can add a simple mapping in my Function rule, to map the json results to my application’s data model. I can then call this function from an automation in my workflow to load the data and use it in my processes or with my agents: 

 

Want to build your own custom integrations faster?  

• Install these skills today in your workspace. 

• Build the custom functionality you need; let your coding agent handle the boilerplate so you can stay focused on your application outcomes. 

• Import your function into your Launchpad application and use it anywhere: your workflows, application logic, and more.