Agents are getting good enough that they are starting to feel like a new kind of tool. Not just for answering questions in a chat, but for helping turn intent into something real.

Software is the obvious example. But I think the more interesting story is what happens when agents start helping with work outside traditional coding.

I’m a 3D printing hobbyist, and I love the idea of designing useful things for the real world: brackets, cable clips, holders, mounts, simple enclosures, print parts for something that broke at your home. The problem is that I’m terrible at 3D modeling. I tried, watched a ton of YouTube tutorials, and still suck at it. I usually know exactly what I want, what needs to fit where, and what constraints matter. I just can’t move from idea to clean 3D model very quickly.

That led me to a simple question: can an agent help me design 3D models that I can actually print on my Bambu 3D printer?

This post is about my journey to answer that question.

Inspired by this CadQuery skill for Claude Code, One of my agents and I created this OpenClaw 3D Printing Skill on ClawHub. The goal was straightforward: give OpenClaw agents a practical workflow for helping design functional mechanical parts in a way that is parametric, repeatable, and easy to go from idea to printing an object.

The skill helps agents do the parts that matter well: gather dimensions and constraints, model parts parametrically in millimeters, export printable formats, and review fit, tolerances, and printability before the next iteration.

That last part is the important one. Most useful 3D printed parts are not right on the first try. You print, test, adjust, and print again. Agents are surprisingly good at that loop when the workflow is clear.

What made this project interesting was not just the modeling side. A big part of the work was making the whole thing practical to install and use, packaging more than a single skill file. I built helper scripts, render and preview tooling, export utilities, example briefs, and setup docs that make the workflow much easier to get running.

It is one thing to make something technically possible just for me. It is another thing to make it usable by other people without unnecessary pain.

I also had to deal with a very real problem that shows up in projects like this: environment setup is often the hardest part. CadQuery installation across different machines took some time, especially if you want something that works well on macOS, Linux, and Raspberry Pi style environments. A lot of the effort went into making that path much clearer and more reliable.

By the end, now there is a public skill repo, packaged the skill, improved the install experience significantly, published it to ClawHub.

I think agents become really powerful when they help close the gap between intent and execution. In this case, I know the problem I want to solve. I know what a good result looks like. I just do not have the fluency in 3D modeling to get there on my own. A good agent workflow can help bridge that gap.

That changes the experience from “I need to become good at this tool before I can build the thing” to “I can describe what I need, iterate with the agent, and get to a usable result much faster.”

To me, that is the cool part. It is so rewarding.

This OpenClaw 3D printing skill is a small project, but it is a concrete example of something I think we will see much more of: agents helping people create real artifacts in domains where they have strong intent and judgment, but less hands-on execution skill.

I hope you can do the same on your area of technical passion.