An end-to-end Codex skill for mathematical modeling competitions such as CUMCM and MCM/ICM. It helps turn a contest problem into a defensible sequence of definition checks, model design, implementation, bounded independent verification, integration, and delivery.
- A reduced stage workflow from whole-problem intake through result delivery
- Answer-validity gates covering problem/model, model/solver, and evidence/claim consistency
- Risk-focused reference modules for information admissibility, geometry, multi-objective models, piecewise rules, spatial non-overlap, and related cases
- A bounded independent-verification protocol that avoids endless audit loops
- Targeted evaluation cases and a dependency-free structure check
Clone this repository into your Codex skills directory:
git clone https://github.com/user0928/math-modeling-contest-workflow.git "$env:USERPROFILE\.codex\skills\math-modeling-contest-workflow"Restart Codex after installation so the skill is discovered.
Invoke the skill explicitly when you want the complete workflow:
Use $math-modeling-contest-workflow to solve this contest problem and keep a reproducible state ledger.
Codex may also select it automatically for end-to-end modeling-contest work. Ordinary single-formula homework and paper-only polishing are intentionally outside its default scope.
SKILL.md Main routing and operating rules
references/stages/ Stage-specific workflow guidance
references/plugins/ Risk-specific checks loaded only when relevant
references/supervision/ Independent verification protocol
evals/ Targeted behavior cases and benchmark summary
scripts/ Dependency-free structural validation
python scripts/verify_skill_structure.pyThe repository also runs this validation automatically on pushes and pull requests.
Keep SKILL.md concise and route conditional detail into focused references. When changing behavior, update the corresponding evaluation case and run the validation script before committing.