Modeling Workbench skill
A skill for mathematical modeling contests: from problem analysis and model comparison to a verified LaTeX paper and PDF.
- 177 ★ on the repository
- Last commit
- MIT
Extended access
Why this level
- The agent runs computation code and check scripts and builds documents in your environment.
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add skills/modeling-workbenchDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- The agent runs computation code and check scripts and builds documents in your environment.
This is third-party code. Review the repository files before installing.
What it does
Modeling Workbench guides a mathematical modeling contestant through seven stages: reading the problem, comparing models, computing, independent review, writing the paper, checking and delivering the PDF. The agent receives the official problem, attachments and current-year requirements, and the human confirms the decision at three points: model selection, conclusions before writing, and final delivery. Every number in the paper is tied to a computation record of code, data and result, and stale evidence is detected after code changes. Models are tried as candidates first and the reason for the choice is recorded. The output is a Chinese or English LaTeX paper, source code with results, review records and a PDF checked page by page. Support starts with CUMCM and extends to MCM/ICM and other contests.
Who it is for. Teams in mathematical modeling contests that need a controlled path from the problem statement to a submitted paper.
Good fit when
- A CUMCM or MCM/ICM problem is released and the team must set up a project with attachments and requirements fast.
- Several models compete and the choice has to be justified with recorded trial runs.
- Numbers in a paper draft must be tied to code and data so they can be rechecked.
- The paper is assembled in Chinese or English and delivered as a PDF checked page by page.
Not a fit when
- You only need ordinary polishing of a finished paper, which the README explicitly excludes.
- You want a guarantee of model validity without your own expertise: the tool organizes the process but judgment stays with the human.
- The environment has no Python 3.10, computation backend or XeLaTeX and cannot be changed.
Example request
Prepare the environment for Modeling Workbench, run `scripts/doctor.py` and list available stages and missing dependencies.Limitations
You need Python 3.10 or newer, a MATLAB or Python computation backend and XeLaTeX with PDF tools. Missing LaTeX does not block analysis and modeling but does block building the PDF. On Windows the author recommends WSL2, and native Windows does not certify full contest delivery. The `.cumcm` workspace format is unchanged, but workspaces older than v0.6 are unsupported. The tool organizes and verifies the process while model suitability and conclusions remain the human's call.
How to disable. Delete the folder with the `modeling-workbench` clone and the projects you created. The skill installs nothing system-wide itself, so remove dependencies added during environment preparation separately.
FAQ
How do I start Modeling Workbench?
First send the agent the environment-preparation prompt from the README: it downloads the repository and runs `scripts/doctor.py`. Then in a new conversation give the tools folder, the official materials folder and a new project folder.
How do I call the skill in Claude Code and Codex?
Type `/cumcm-workflow` in Claude Code and `$cumcm-workflow` in Codex. The skill name and workspace format stayed the same even though the repository is named `modeling-workbench`.
Where does the human confirm the work?
At three points: model selection, conclusions before writing the paper and final delivery. The agent waits for explicit confirmation at each.
How do I resume an interrupted project?
Give the project folder and ask to resume `cumcm-workflow`. There is no need to initialize again: the skill picks up the current stage.
Can the environment be checked without installing?
Yes. The `doctor.py` and `project_status.py` scripts only report missing dependencies, evidence and pending decisions and do not install anything or change project state.
Discussion
No questions or reviews yet.
Related
SkillsA skill for mathematical modeling contests: from problem analysis to verified results, with a workbench for a three-person team.
A meta-skill that splits a task into stages, picks a profile and runs multi-agent work in Codex and Claude Code.
brainstorm-mcp runs debates among GPT, Gemini, DeepSeek, Claude and Ollama models in your editor and returns a three-point verdict
A Claude Code plugin that asks several AI models the same question and shows their answers side by side with a synthesis of agreement and divergence
Скилл ведет читателя по научной статье в шесть шагов, присылая нужные рисунки: от мотивации до разбора примера по коду.
A meta-skill that distills methodologies from books, long videos, podcasts and courses into a set of ready-to-invoke agent skills
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Repository stars | Main difference |
|---|---|---|---|
| Modeling Workbench (this card) | no | 177 | |
| Math Modeling Skill | no | 1,929 | a single modeling skill; Modeling Workbench runs the whole cycle with computation records and three checkpoints |
| MathModeling-skills | no | 1,091 | a set of modeling skills; this one adds end-to-end evidence checking and PDF delivery |
README in short
The repository grew out of the CUMCM workflow and the shared pipeline of computation, review and delivery serves other contests too. Startup is described through two prompts to the agent.