I kept learning from talks and articles but losing the details. This turns source material into a skill you can talk to.
I'd take notes, highlight transcripts, save articles — and still lose the thread a week later. The signal was always structural, not quotable. Principles, mental models, decision trees. I needed a way to preserve that structure, not the paragraphs.
Instead of summarizing top-down, extract bottom-up: what are the author's principles, their mental models, the branches in their decision-making? A skill that encodes those so I can interrogate them later.
Skill Distillery is the skill I use to create every other skill on this page. Point it at a transcript, an article, or a framework. It distills the thinking, not the prose. Every new skill downstream of this one got faster to build.
I spent months in the habit of devouring design talks, founder essays, and framework articles — and watching most of them evaporate by the following week. The people I admired most were telling me how they thought, and somehow the thinking wasn’t sticking.
What I realized: the part I needed to keep wasn’t the paragraphs. It was the structure underneath — the principles they’d repeat across contexts, the mental models they’d built up, the decision forks they’d been through. That structure was what I wanted to come back to.
Skill Distillery reads a source — article, transcript, framework, book chapter — and works bottom-up. It extracts principles in the author’s voice, the mental models they rely on, the trees of decisions they walk through. The output is a SKILL.md that you can talk to the way you’d talk to the author at a coffee shop. Interrogate it. Ask how they’d apply a principle to your exact problem. Push on an edge case.
It’s the skill that builds every other skill on this page. Point it at a talk; ship the skill by the end of the afternoon.
Install via Claude Code’s plugin system. Source transcripts or article URLs as arguments; the distillery does the rest.
$claude plugin install kylezantos/skill-distillery