| name | modeling-paperkit |
| description | Use this skill when working with the ModelingPaperKit repository to write, edit, compile, review, or prepare math modeling contest papers for CUMCM, MCM/ICM, Wuyi, Beijing, or the included example templates. Use it for LaTeX template edits, section organization, figure/table insertion, build-log diagnosis, final submission checks, contest safety constraints, AI usage/source logging, and PaperKit-specific workflows. |
| metadata | {"short-description":"Work on ModelingPaperKit math modeling papers"} |
ModelingPaperKit
Use this skill for math modeling contest paper delivery inside the ModelingPaperKit repository. The job is to help with structure, LaTeX, build/review loops, contest-specific checks, and safe handoff to deterministic scripts.
Do not treat this as an automatic contest-solving skill. Do not fabricate data, model results, citations, figures, awards, or submission readiness.
First Steps
- Identify the target:
cumcm, mcm, wuyi, beijing, or example.
- Inspect the relevant main file and sections before editing.
- Load only the reference files needed for the task:
- Repository layout:
references/repository-map.md
- Standard workflow:
references/workflow.md
- LaTeX macros and local conventions:
references/latex-patterns.md
- Final review:
references/final-review.md
- Contest safety:
references/contest-safety.md
- CUMCM:
references/cumcm.md
- MCM/ICM:
references/mcm.md
- Wuyi:
references/wuyi.md
- Beijing:
references/beijing.md
- Prefer existing repository scripts over ad hoc shell logic:
- Build:
python scripts/build.py --target <target>
- Full verification:
python scripts/verify_build.py
- Clean:
python scripts/clean.py
- New contest scaffold:
python scripts/new_contest.py --name <name> --lang zh|en
Workflow
For writing or editing:
- Read the target
main_*.tex.
- Read only the relevant
sections/*.tex.
- Use local PaperKit macros and style conventions.
- Keep edits scoped to user-requested sections.
- Build or run checks when the user asks for verification or when edits may break LaTeX.
For build failures:
- Run the target build if needed.
- Read the corresponding
out/*.log.
- Summarize the first actionable errors.
- Make the smallest safe LaTeX fix.
- Rebuild and report remaining issues.
For final review:
- Check build status and PDF presence.
- Check labels, references, figures, tables, and special pages.
- Check suspected identity leaks.
- Check contest-specific AI/source/supporting-material requirements.
- Report findings as
Critical, Warning, or Info.
Contest Safety
If the user is actively competing:
- Ask or infer whether the mode is
cumcm_active, mcm_active, or off.
- In
cumcm_active, do not proactively search public discussion platforms or solution repositories for current contest problems.
- In
mcm_active, public inanimate sources may be used only with source logging and citation discipline.
- Never help submit files automatically.
Read references/contest-safety.md when current-contest rules, browsing, AI use, or source logging matters.
Output Style
When reporting review results, lead with findings:
Critical
- ...
Warning
- ...
Info
- ...
When making edits, mention the files changed and the verification command run. If tests or builds were not run, say so clearly.