Week 2 Theme: Tooling for AI Tools
Spring into AI Competition Week 2
Build something designed for AI to use. When the user is an agent, not a human, everything changes-or does it? What if the software humans write in the future isn’t for… humans per se, but for agents? After all, there’s only 8 or so billion humans. There’s going to be a lot more software agents. So let’s build tools for the future.
AI Skill Tools
The things you can do with AI skills are jawdropping. Take five minutes and read this website https://agentskills.io/home as the return on that will be a significant unlock for you. Jump down to the video for a recipe that’ll show you how to use Copilot to build a tool.
It’s 5:30 am and I hear someone ask, “Do you have other ideas for me to consider?” Yes-and also I appear to be hallucinating voices. That was actually the coffee pot.
I’ve built out this page to get you started thinking about this theme. It’ll have a bunch of ideas. Your submission can be a standalone tool-so a tool that produces say game art, or an API integration and so on. The idea is you’re building a useful tool that an AI system can use. You have a simple AI system - Copilot and Codex. If you want to build for OpenClaw, well, that is also an AI system. It’s a pretty tolerant theme.
The first time I built an AI skill was a material unlock in what I could do. If you’re wondering what’s one simple thing you can do right now that’d completely change what you can do with AI? It’s tooling for AI tools.
Where humans typically get tripped up at is the process. I did. So I have something for you.
See, I cover a real simple example about the process of tool building for AI using AI tools here. That’s right- using AI to build a tool that AI uses. You don’t need to watch this video, just search for videos of people explaining claude skills.
Video summary:
In this video, Eric from Advisory Hour introduces his new YouTube channel, explaining its origin in a public speaking book that encourages embracing embarrassment. He aims to bridge the gap between rapidly evolving AI technology and real-world applications by chronicling his learning process.
Eric demonstrates building an Agent Skill in Visual Studio Code. He walks through the basic structure of Agent Skills, including the skill.md file and scripts folder. He then attempts to create an RPG character generator skill using a Python script, encountering challenges due to a missing Python environment and conflicts with an existing MCP server.
To resolve these issues, Eric converts the Python script to PowerShell with the help of AI and debugs the skill, demonstrating how the AI can assist in correcting errors . He concludes by reflecting on the learning process, the importance of trying new things, and the need to find practical applications for AI technology.
So there’s some laughs, some learnings and some tears with python in that video.
No matter which path you take, there’s one thing for certain.
Week 2 has begun.
Good luck!