← ClaudeAtlas

explainer-intuitionlisted

Explain the underlying mental model of a topic — why it works the way it does, not just what it is. Use when the user runs /explainer-intuition, or asks "help me get an intuition for", "why does this actually work", "what's really going on under the hood", or "I know the definition but it doesn't click".
jbordoe/claude-skills · ★ 0 · AI & Automation · score 53
Install: claude install-skill jbordoe/claude-skills
Give the reader the *mental model* behind a topic — the picture an expert holds in their head that makes the whole thing feel obvious. The reader may already know the definition or the mechanics; what's missing is the "of course" feeling. Your job is to install that. This is a different axis from ELI5. ELI5 makes a topic *simple*. Intuition makes it *click* — it can assume more, and it goes for the load-bearing "why", not the gentle "what". ## Principles - **Find the one key idea.** Most topics have a single insight that, once seen, makes everything else follow. Lead with it. State it plainly, early. - **Explain why it MUST be this way.** Don't just describe the design — show the forces that make it the natural answer. "Given X, you'd be forced to invent Y" beats "Y is defined as…". - **Reason from the constraints.** What problem is this solving? What goes wrong with the obvious naive approach? The intuition usually lives in *why the naive version fails*. - **Anchor with a vivid model.** A physical picture, a spatial metaphor, a limiting case, a "what if you pushed it to the extreme". Make it something the reader can *run* in their head. - **Use limiting / edge cases as intuition pumps.** "What happens when N=1? When N→∞?" Extremes often expose the mechanism the middle hides. - **Connect to something they already grok.** The best intuition is "it's the same shape as this other thing you already understand." Name the isomorphism. ## Anti-patterns - Do not just restate the