Mahjong Skill
A small, half-serious skill project for mahjong-flavored reasoning
Mahjong Skill is a small side project that started closer to a joke than a product plan.
The idea is simple: package mahjong-related habits, terminology, and decision patterns into a reusable skill. It is not trying to become a professional mahjong engine. It is more like giving an AI assistant a tiny corner of table culture: enough structure to talk about tiles, hands, risk, and questionable decisions with a little more context than a generic chatbot.
That is the fun part of skills. They do not always need to be huge systems. Sometimes a skill is valuable because it captures a weird, specific slice of knowledge and makes it easy to reuse.
Why I Built It
I like small projects that are allowed to be a little silly.
Mahjong is full of local language, table intuition, half-math, and emotional damage. That makes it a good playground for skill design. If a skill can describe the domain clearly, keep the tone right, and avoid pretending to be more certain than it is, then the same pattern can work for more serious domains too.
This project is mostly a playful experiment, but it still asks a useful engineering question:
How much domain behavior can be captured in a lightweight skill before you need a full application?
What Makes It Interesting
The project sits in a nice middle ground. It is small enough to stay fun, but specific enough to be a real test of instruction design.
A generic assistant can explain mahjong rules. A skill should do something more opinionated. It should know what kind of framing is useful, what details matter, and when to avoid overclaiming. In other words, the project is less about "AI plays mahjong" and more about "AI talks about mahjong in a way that feels less lost."
That distinction matters. A good skill is not just a pile of facts. It is a reusable behavior package.