# AI use cases for Modeling Workbench (Lucasuiii/modeling-workbench)

Repo: [Lucasuiii/modeling-workbench](https://github.com/Lucasuiii/modeling-workbench) | Use cases: 4 | Page: https://dedupedai.com/use-cases/lucasuiii-modeling-workbench/

A Codex and Claude Code workflow that guides math modeling contest work from official problem files to a checked, bilingual LaTeX and PDF paper, with evidence records for each step.

This file holds every use case of this repo. Each one starts with a "## " heading and ends with its URL.

- Check which modeling stages this machine can run: https://dedupedai.com/use-cases/lucasuiii-modeling-workbench/check-modeling-environment/
- Start a contest project from official problem files: https://dedupedai.com/use-cases/lucasuiii-modeling-workbench/start-contest-project-from-official-files/
- Test two candidate models before choosing one: https://dedupedai.com/use-cases/lucasuiii-modeling-workbench/compare-two-candidate-models/
- See where a modeling project stands and what needs your sign-off: https://dedupedai.com/use-cases/lucasuiii-modeling-workbench/review-project-status-and-gates/

## How to check which modeling stages this machine can run with Modeling Workbench

Repo: [Lucasuiii/modeling-workbench](https://github.com/Lucasuiii/modeling-workbench) | Difficulty: Intermediate | Time: 1 hour 30 minutes | Basis: From the README

Run the doctor check to see which stages work and which dependencies are missing.

**Who it is for:** A student preparing for a math modeling contest who wants to know whether Python, LaTeX and the calculation backend are ready before the contest starts.

**The problem:** Setting up the toolchain by guesswork wastes time, and missing tools are easy to miss until the paper stage.

**The outcome:** A report of usable stages and missing dependencies, with no software installed or changed unless you agree.

### Steps

1. Put the repository in a separate tool folder, or unzip the Download ZIP from the repository page into one.
2. Open Codex or Claude Code in a writable local folder and send the environment prompt from the README Quick Start.
3. Set S to the scripts folder inside .agents/skills/cumcm-workflow/scripts, then run `python3 "$S/doctor.py"`.
4. Add --json to the same command to see the full structured result.
5. Read which stages are available and which dependencies are missing, then decide what to install.

### Prompt to paste into your agent

```text
Use the Modeling Workbench repo in <tool folder> to check my environment. Read .agents/skills/cumcm-workflow/SKILL.md, run scripts/doctor.py, and list the available stages and missing dependencies. Do not install or change anything without my approval, and confirm the report names the tool folder and version.
```

### Caveat

Setup time depends on existing dependencies, and the first LaTeX setup can take a long time.

Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/

URL: https://dedupedai.com/use-cases/lucasuiii-modeling-workbench/check-modeling-environment/

## How to start a contest project from official problem files with Modeling Workbench

Repo: [Lucasuiii/modeling-workbench](https://github.com/Lucasuiii/modeling-workbench) | Difficulty: Intermediate | Time: 2 hours | Basis: From the README

Copy the official problem files into a new project folder with fixed file identities.

**Who it is for:** A modeling team that has just received a contest problem and its attachments and needs a clean project to work in.

**The problem:** Working directly in the original problem folder risks losing track of which files were official and what changed.

**The outcome:** A new project folder with the official files copied in and recorded, while the original folder stays unchanged.

### Steps

1. Create a new output folder that does not exist yet, separate from the tool and material folders.
2. Place the official problem files, attachments and any format files in one material folder.
3. In a new chat, send the project start prompt from the README Quick Start and replace the three paths.
4. When asked, confirm the contest, year, paper language and submission rules.
5. Review the problem facts and the sub-question capabilities the agent writes before starting model design. (inferred, check the README)

### Prompt to paste into your agent

```text
Use the Modeling Workbench repo in <tool folder> to start a new contest project. Use <official material folder> as the official material and <new project folder> as the output folder. Confirm the contest name, year and paper language, then report the copied file list and the stage that comes next.
```

### Caveat

Contest-specific template and submission rules still need to be checked against the official documents by hand.

Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/

URL: https://dedupedai.com/use-cases/lucasuiii-modeling-workbench/start-contest-project-from-official-files/

## How to test two candidate models before choosing one with Modeling Workbench

Repo: [Lucasuiii/modeling-workbench](https://github.com/Lucasuiii/modeling-workbench) | Difficulty: Advanced | Time: 2 hours 30 minutes | Basis: From the README

Run both candidates as cheap exploratory runs and record why one was selected.

**Who it is for:** A modeler on a contest team who has two plausible approaches to one problem and must justify the choice in the paper.

**The problem:** Picking a model first and testing it later makes it hard to show a real comparison or explain the choice.

**The outcome:** A recorded comparison table of candidates, with exploratory run evidence and a written reason for the selected model.

### Steps

1. In MODEL_CONTRACT.json, write two candidates, each with why_considered and discriminating_evidence.
2. Run each candidate as an exploratory run: `python3 "$S/record_run.py" --project <project folder> --candidate CAND-A -- python3 code/try_a.py`. (inferred, check the README)
3. Repeat the same command for CAND-B with its own script.
4. Mark the chosen candidate as selected and write decision_rationale that cites the exploratory run IDs.
5. Run `python3 "$S/cumcm_check.py" --project <project folder> --stage model-design --gate-mode preflight` to print the candidate comparison table. (inferred, check the README)

### Prompt to paste into your agent

```text
Use the Modeling Workbench repo in <tool folder> to compare two candidate models for the problem in <project folder>. Record an exploratory run for each candidate, select one with a written rationale, and run the model-design check. Verify the comparison table lists both candidates and the selected one cites its runs.
```

### Caveat

The workflow records and checks the comparison, but it does not prove that the chosen model is mathematically correct or globally optimal.

Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/

URL: https://dedupedai.com/use-cases/lucasuiii-modeling-workbench/compare-two-candidate-models/

## How to see where a modeling project stands and what needs your sign-off with Modeling Workbench

Repo: [Lucasuiii/modeling-workbench](https://github.com/Lucasuiii/modeling-workbench) | Difficulty: Intermediate | Time: 30 minutes | Basis: From the README

Get a status report of stages, evidence problems and pending confirmations.

**Who it is for:** A team lead on a modeling contest who needs to know what is finished, what is blocked, and what still needs a human decision.

**The problem:** A long project spread over several sessions is hard to resume without a clear list of open items.

**The outcome:** A status report of the current stage, evidence issues and the decisions waiting for explicit confirmation.

### Steps

1. Run `python3 "$S/project_status.py" --project <project folder>` with S set to the scripts folder. (inferred, check the README)
2. Add --json to see the full structured result.
3. Read the evidence problems and the pending confirmations the report lists.
4. Show the current material to the team, then record the decision after an explicit reply, using record_decision.py with --confirm-human.
5. Repeat at model selection, before paper writing, and at final delivery, which are the three required confirmation points.

### Prompt to paste into your agent

```text
Use the Modeling Workbench repo in <tool folder> to check the status of the project in <project folder>. Run the project status tool, list the evidence problems and pending confirmations, and do not record any decision unless I confirm it in my own words.
```

### Caveat

Status checks report what the tool can verify; a passing check does not replace reading the model and the results yourself.

Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/

URL: https://dedupedai.com/use-cases/lucasuiii-modeling-workbench/review-project-status-and-gates/