dsh-math-modeling-skills-Gatecraft-

A familiar frustration in math-modeling competitions: the model kernel works, yet points leak in the expression and acceptance layers. Figure overflow, dual-track numbering, AI tool names in references, a magnitude jump in the abstract individually minor, collectively the gap between a national award and a provincial one. Worse, existing automated checks are almost entirely tied to the LaTeX pipeline, while the docx artifact that actually gets submitted silently bypasses them and the mess is only caught at the last minute.

GateCraft packages nine skills and a full asset set into a gate-controlled pipeline: five EDA questions, a three-part verification kit, and scripted quality checks form the checkpoints. Before delivery, docgate.py runs thirteen mechanical checks against both docx (OOXML) and tex backends; any unresolved FAIL blocks submission. External solvers such as MathModelAgent can be plugged in as backends while GateCraft stays focused on orchestration, pivoting, and deep QC thinking happens at the gates, not behind a one-click end-to-end run.

Best suited for teams chasing national-level awards on statistical-analysis and optimization/decision problems (typical C type). If your model already runs but you keep losing points on formatting, phrasing, and acceptance details, this gate-craft discipline is built for you. Mechanism simulation and graph-theory/engineering problems (A and B types) remain untested in practice; extend the checklist and give back to the community.