MathModeling-skills

Why we included this project

Teams in math-modeling contests usually lose the most time to consistency, not to learning models: every number in the paper has to trace back to a frozen snapshot of the data and code, and that has to hold under deadline pressure. This project packages that discipline into a set of skills for Claude Code and Codex, where the assistant profiles data, runs method-specific risk checks, verifies figures render correctly, and audits consistency, while the human keeps the modeling judgment. The design is deliberately conservative about who decides what. Two of the hard gates are the user's call, not the AI's, and a three-auditor layer has the final say on whether the paper is ready to submit. Teams competing in CUMCM or MCM/ICM who want an assistant that enforces rigor instead of ghost-writing the whole entry will find it worth a look.