Top AI Repos — open-source AI, indexed and scored
Top AI Repos tracks AI repositories on GitHub and answers two different questions about each one: is it moving right now, and would you bet a product on it.
Top AI Repos tracks AI repositories on GitHub and answers two different questions about each one: is it moving right now, and would you bet a product on it.
面向数学建模竞赛的 Claude Code / Codex Skills ，支持分阶段建模流程与 Python、MATLAB/北太天元代码分支。
| Date | Stars |
|---|---|
| 2026-07-31 | 365 |
| 2026-08-01 | 376 |
| 2026-08-11 | 470 |
| 2026-08-18 | 542 |
| 2026-08-19 | 551 |
| 2026-08-20 | 563 |
| 2026-08-21 | 566 |
| 2026-08-22 | 567 |
| 2026-08-23 | 573 |
| 2026-08-24 | 585 |
| 2026-08-25 | 599 |
| 2026-08-26 | 610 |
| 2026-08-27 | 618 |
| 2026-08-28 | 634 |
| 2026-08-29 | 644 |
| 2026-08-30 | 657 |
| 2026-08-31 | 668 |
| 2026-09-01 | 680 |
| 2026-09-02 | 692 |
| 2026-09-03 | 701 |
| 2026-09-04 | 710 |
| 2026-09-05 | 723 |
| 2026-09-06 | 740 |
| 2026-09-07 | 766 |
| 2026-09-08 | 792 |
| 2026-09-09 | 814 |
| 2026-09-10 | 865 |
| 2026-09-11 | 879 |
| 2026-09-12 | 888 |
| 2026-09-13 | 889 |
| 2026-09-14 | 890 |
| 2026-09-15 | 892 |
| 2026-09-16 | 902 |
| 2026-09-17 | 914 |
| 2026-09-18 | 924 |
| 2026-09-19 | 933 |
| 2026-09-20 | 951 |
Today
+18 stars today
This week
+62 stars this week
This month
+385 stars this month
Momentum
134.0
growth rate 6.97%/day
<p align="center"> <img src="docs/assets/logo.svg" alt="MathModeling-skills" width="640"/> </p> <p align="center"> <a href="./README.md"><b>English</b></a> · <a href="./README-zh.md">简体中文</a> · <a href="./CLAUDE.md">Project Rules</a> · <a href="./Initial%20Prompt.md">Initial Prompt</a> · <a href="mailto:[email protected]">📧 Contact</a> </p> <p align="center"> <img alt="License" src="https://img.shields.io/badge/license-MIT-2E9E44"> <img alt="Skills" src="https://img.shields.io/badge/skills-28-1A6FC4"> <img alt="Claude Code" src="https://img.shields.io/badge/Claude%20Code-supported-E28E2C"> <img alt="Codex" src="https://img.shields.io/badge/Codex-supported-E28E2C"> </p> --- > [!NOTE] > **Update — this is now an assistant, not an autopilot.** The earlier version ran the whole contest end to end and left the user only clicking "confirm", which is closer to ghost-writing: it does not fit most contests' rules, and it does little for your own skills. This version hands the key judgments back to the user — the AI returns to a supporting role, and you stay in charge. The skill count went from 24 to 28. The previous full-auto version is kept intact on the [**`legacy-full-auto`**](https://github.com/zhnnky329/MathModeling-skills/tree/legacy-full-auto) branch; switch to it if you prefer the old behavior. > A set of skills for math-modeling contests, built around the mistakes that tend to cost the most time. They sit behind a set of hard gates — two of which the user decides, not the AI — and a three-auditor layer that has the final say on whether the paper is ready to submit. The aim is not to automate more, but to make sure no step can quietly skip a check: every number in the paper traces back to a frozen snapshot, every reviewer leaves a file on disk, and no skill marks itself as "done". > > The principle it is built on: **the AI owns mechanical correctness; the user owns modeling judgment.** It profiles data, runs method-specific risk probes, freezes numbers, render-checks figures, and audits consistency. It does not choose the method, decide what a number means, or invent the reasoning behind a choice. > > Found a bug, or want to share how it went in a real contest? Email **[[email protected]](mailto:[email protected])**, or open an issue. ## Why this exists When a team loses a modeling contest, it is rarely because they did not know enough models. It is usually one of these: - They misread what the problem was actually asking. - They skipped the baseline and went straight to a complex model that nobody could explain later. - The paper states a number that no script in the repo actually produces. - A bug gets fixed late in the process, but the paper still carries the numbers from before the fix. These are workflow problems, not modeling problems. The skills here are arranged to make these failures hard to hide. ## What's different | | A typical pipeline | This one | |---|---|---| | How you move on | "this stage is done, next" | Each gate has an explicit pass condition. Fail it and everything downstream gets marked stale. | | Which method, and why | The AI picks and writes the justification | You choose the trade-off; the AI screens a main candidate, a usable baseline, and at most one conditional fallback; **you** commit the route and reason (Gate G2.5). | | From idea to code | A method is accepted if the math looks right | A time-bounded risk probe checks data coverage, assumptions, output degeneracy, perturbation sensitivity, and scale (Gate G2). | | Code review | Someone says "looks fine" | A compact JSON review must pass named syntax, input, method-alignment, reproducibility, and output checks (Gate G3). | | Numbers in the paper | Re-read from the latest results each time | Frozen into `frozen_numbers.json`. Changing one means logging the change and re-freezing (Gate G4) | | Exploration cost | Full reports and audits at every step | `lean` keeps manifests, decisions, probes, and run sum
Excerpt of 18,532 characters
Read on GitHub
35
Would you bet a product on this? Bounded 0–100 and slow moving.
matched fp:23c00ed9babc45be, llm:description: '面向数学建模竞赛的 Claude Code / Codex Skills ，支持分阶段建模流程与 Python、MATLAB/北太天元代码分支.' (Chinese)
matched fp:23c00ed9babc45be, llm:description: '面向数学建模竞赛的 Claude Code / Codex Skills ，支持分阶段建模流程与 Python、MATLAB/北太天元代码分支.' (Chinese)
matched fp:23c00ed9babc45be, llm:description: '面向数学建模竞赛的 Claude Code / Codex Skills ，支持分阶段建模流程与 Python、MATLAB/北太天元代码分支.' (Chinese)