woodfishhhh/ez_math_model

EZ Math Model is a Chinese-language Agent Skill pack for math modeling competitions. It is built for events like China's CUMCM, the MCM and ICM contests in the United States, and similar graduate-level modeling competitions, as well as classroom modeling assignments. The user hands the contest problem and any data files to an AI agent, and the skill walks the agent through a fixed pipeline that ends with a finished paper and a delivery zip. Once installed, the skill is run by adding it to either Codex or Claude Code, then asking the agent to use it on a particular problem. The repository shows the install commands for both agents, and the user is expected to put the problem statement, any extra notes, and data attachments into a folder inside a project directory. The first run must complete a setup step, and any tool decisions the user has not confirmed are kept temporary rather than written as permanent defaults. The fixed pipeline has seven stages. Stage 00 sets up the environment and runtime folder. Stage 01 reads the problem and saves files like problem.md and intake.json. Stage 02 picks the modeling approach. Stage 03 writes and runs Python code, then exports results, figures, and a chart manifest. Stage 04 writes the paper, stage 05 audits quality, and stage 06 packages the output. The final delivery is a paper in four formats, Markdown, DOCX, plain text, and PDF, along with a zip of the whole project folder. The skill ships with a built-in library of algorithms for optimization, prediction, evaluation, graph theory, statistics, machine learning, and combined methods. It also bundles sub-skills for reading PDFs, Word files, and spreadsheets, for paper search and web context, for dataset discovery, and for brainstorming, debugging, and result polishing. The repository is MIT licensed, includes an English README and a star map of capabilities, and points to a QQ group for support.