A [Claude Skill](https://docs.claude.com/en/docs/agents-and-tools/agent-skills/overview) that gives algorithm work a **fixed, predictable structure** — whether you are learning a new problem, debugging your own code, planning what to practice, or proving something for a class.

This is not a tool for solving more problems; Claude can already do that. It is a tool for making the *response* consistent: the same sections, in the same order, with the same things pinned down, every time.

Available in **English** and **Chinese (中文)**. See [Install](#install).

---

The problem it solves

Ask an LLM about an algorithm problem twice and you get two differently-shaped answers. One time it opens with code, another time with a long analogy. Sometimes it walks through an example you did not need; sometimes it skips the boundary condition you were actually stuck on. When you are working through a few hundred problems, that inconsistency is the friction.

This skill fixes the shape of the answer so you always know where to look — and, just as importantly, defines when a section should be **left out** rather than padded.

---

Design principles

Seven decisions shape everything else here. They are worth stating because several of them cut against what a prompt of this kind usually does.

**1. Structure, not knowledge.** The skill does not make the model better at algorithms — nothing in a prompt can. It makes the *shape* of the answer predictable, so you always know where the complexity analysis is and where the "here is how people get this wrong" note is.

**2. Omission is a feature.** Several sections are conditional, each with a written test. `Underlying Idea` appears only if it passes *could this paragraph guide you through a problem you have not seen yet?* — a takeaway that merely restates the solution gets dropped. The instruction throughout