Last Summer, I Watched a Glassblower at Work—Then Found His Craft on GitHub
Last summer, I was wandering through a shopping mall when I spotted a glass studio doing a live demonstration. The craftsman held a long blowpipe, scooped a glowing blob of molten glass from the furnace, and blew while spinning it with movements so fluid they looked almost effortless. In less than ten minutes, a delicate little vase appeared before my eyes. I stood there behind the glass window watching for a long time, thinking to myself: if only more people could learn this craft.
Later, while browsing GitHub, I stumbled upon a project called domain-experts. Its tagline caught my attention—"Turning human experts into AI agents." What surprised me even more was that it actually contained a skill file specifically for the trade of "glass blowing and mold forming," located at the path roles/glass-blower-molder. I was taken aback. Glass craft is something that depends heavily on hand-feel, experience, and on-the-spot judgment, isn't it? Can that really be turned into an AI skill?
What Does This Skill Actually Do?
In simple terms, this skill takes the hard-won experience stored in the minds of master glassblowers and organizes it into a knowledge structure that AI can understand and invoke. It's not a video tutorial or an operation manual. Rather, it's a role definition + knowledge framework that allows AI to genuinely step into the mindset of a "glass artisan" when conversing with users—thinking and answering questions from that perspective.
For example, if you ask it, "When blowing a 20-centimeter-tall vase, roughly what temperature should the molten glass be controlled at?" it will give you a credible reference based on thermal principles, rather than scraping some random passage off the internet to fob you off. If you want to understand the actual differences between mold-formed and free-blown pieces in terms of texture, luster, and detail, it can explain that clearly from the perspective of process principles as well. This kind of experience is completely different from searching for information on a search engine—it feels more like chatting with a knowledgeable master craftsman.
I took the time to dig into the repository's structure. The glass-blowing role directory mainly contains two things: a SKILL.md file and a references folder. SKILL.md is the core of the entire skill, defining the role's positioning, capability boundaries, and conversation strategies. The references folder presumably holds in-depth reference materials that the AI only consults when the conversation touches on relevant topics. This design approach is actually quite sound, because the body of glass craft knowledge is extremely fragmented—cramming everything in at once would actually degrade the quality of responses.
Is Installation a Hassle? It's Actually a Single Command
The installation process was far simpler than I expected. The only prerequisite is that you have a Node.js environment on your computer. Then you just type this one line into your terminal:
npx skills add https://github.com/wonsukchoi/domain-experts.git --skill glass-blower-molderThe way it works is that it pulls down the entire domain-experts repository and then extracts only the glass-blower-molder role skill to install locally. Once installed, you can load this role in whatever AI tool you regularly use and have it converse with you as a glass craft expert.
My Actual Experience Using It
I genuinely installed it and gave it a try, asking questions about the coefficient of thermal expansion of glass, annealing process curves, mold material selection, and similar topics. To be honest, the quality of the answers exceeded my expectations. At the very least, you can tell that it actually understands the fundamental logic of glass craft—it's not just making things up to go along with whatever you say. Of course, if you compare it to a master craftsman with decades of experience, there's still a significant gap. But as a consultant that's available online anytime, ready whenever you call, I'd say it's more than adequate.
Interestingly, this project is still in its early stages. On GitHub, it has just 10 stars and 1 fork, and there's one pending issue in the Issues section. That said, I find projects like this—small but specialized—quite rare and valuable. The author, wonsukchoi, appears to be genuinely and seriously collecting expert knowledge across various domains, not just putting together a gimmick. If you happen to notice obvious inaccuracies in its answers to certain questions, you can absolutely file an Issue or submit a Pull Request in the repository. That's what open-source projects are all about—they're nurtured by the community.
Who Would Find This Skill Useful?
After thinking it over, I figure there are a few types of people who can get real value from it:
- Glass craft enthusiasts—people who want to learn about blowing principles and material properties but can't find a reliable source of information. They can just ask the AI directly, saving both time and effort.
- Design students—when designing glass products, they need to reference forming processes and mold solutions, and this role can offer plenty of inspiration.
- Manufacturing professionals—those dealing with process parameter adjustments in industrial glass mold forming can use it as a supplementary reference.
- Simply curious onlookers—people like me, haha. In just a few minutes, you can gain a basic understanding of this craft.
I've also noticed a broader trend. Projects like domain-experts are, at their core, turning human knowledge into assets. In the old days, an apprentice would spend three to five years under a master before becoming qualified. Now, by distilling experience into structured skill files, AI can "replicate" these experts' judgment at scale. While we're still far from truly replacing human experts, it at least lowers the barrier to knowledge acquisition—and that in itself is something of great value.
One final note: if you're a complete beginner in glass craft like me, don't jump straight into highly advanced questions the first time you use this skill. Start with the basics—raw materials, temperature, and other fundamental concepts—and you'll gradually discover just how much it actually knows. At the end of the day, the idea of packing a glassblower's craft into an AI is, in itself, a rather romantic notion...
