mathmodel-skill

Managing an end-to-end math modeling competition workflow requires coordinating stages from problem selection to final review. This skill provides a 10-stage guided workflow with interactive decision points and competition-specific templates to ensure consistent, reproducible outcomes.

Yes, you can use this math modeling workflow across different harnesses. It features harness-agnostic state persistence that saves progress in a cross-harness codified state file, allowing teams to switch between Codex and Claude Code without losing context.

The math modeling workflow supports CUMCM, MCM/ICM, and Diangong Cup competitions. It applies a fixed sequencing of decision points with competition-specific templates and rubric overlays to drive repeatable results tailored to each contest.

To start the math modeling workflow, you answer five initial questions to initialize the state and begin Stage 0 kickoff. This frontmatter-driven startup loads competition-specific seeds and templates to drive the pipeline.

Yes, the math modeling workflow supports reproducible results through patch-based refinements. It enforces frontmatter-driven startup, elastic templates, and provides optional scripts to automate scoring, patching, and document preparation.

State persistence for math modeling contests relies on a codified state file to maintain progress across harnesses. While it supports patch-based refinements and reproducible results, teams must strictly follow the 10-stage fixed sequencing and frontmatter-driven startup to avoid breaking the workflow.