What is this?
This is an Agent Skill that provides a complete step-by-step review workflow for writing a CUMCM (China Undergraduate Mathematical Contest in Modeling) paper, compatible with Codex, Claude Code and other agents that support the Agent Skills spec.
This is an Agent Skill that provides a complete step-by-step review workflow for writing a CUMCM (China Undergraduate Mathematical Contest in Modeling) paper, compatible with Codex, Claude Code and other agents that support the Agent Skills spec.Is this a standalone app or just a skill file?It literally is the skill: the repository root contains SKILL.md, and you just clone or copy the whole folder into your agent's skills directory - there is no separate build or install step.
It literally is the skill: the repository root contains SKILL.md, and you just clone or copy the whole folder into your agent's skills directory - there is no separate build or install step.How do I get it running with my agent?On Windows, copy the whole folder into the agent's skills directory (e.g. C:\Users\ \.codex\skills\); on macOS/Linux copy it to the equivalent skills path and restart the agent so it shows up in the skill list. Then just tell the agent to use 'cumcm-step-review' on your contest problem, or paste the problem and attachments directly.
On Windows, copy the whole folder into the agent's skills directory (e.g. C:\Users\ \.codex\skills\); on macOS/Linux copy it to the equivalent skills path and restart the agent so it shows up in the skill list. Then just tell the agent to use 'cumcm-step-review' on your contest problem, or paste the problem and attachments directly.Which coding agents actually support this skill?Any agent implementing the Agent Skills specification can load it; the README explicitly confirms Codex and Claude Code as compatible hosts.
Any agent implementing the Agent Skills specification can load it; the README explicitly confirms Codex and Claude Code as compatible hosts.Does it enforce a fixed official paper template?No - the actual paper structure and formatting still follow whatever the current year's official contest rules and template say. The skill only drives the review process and must have core modeling and problem analysis led and manually checked by the actual contest team for compliance.
No - the actual paper structure and formatting still follow whatever the current year's official contest rules and template say. The skill only drives the review process and must have core modeling and problem analysis led and manually checked by the actual contest team for compliance.What does the step-by-step review actually cover?It walks through every section of a modeling paper in order - problem restatement, data preprocessing, per-question modeling and solving, model analysis and testing, evaluation and extension, references, appendix, assumptions, notation, a two-phase abstract, and a final AI usage report - requiring user approval at each step before moving on.
It walks through every section of a modeling paper in order - problem restatement, data preprocessing, per-question modeling and solving, model analysis and testing, evaluation and extension, references, appendix, assumptions, notation, a two-phase abstract, and a final AI usage report - requiring user approval at each step before moving on.How does it keep the figures publication-quality?It enforces hard chart rules: no in-figure titles, captions below the figure, analysis text right after each one, max 3 figures of the same type, print-ready font sizes, a single 300 DPI PNG output, and 7 journal color palettes like Nature, Science, Cell and IEEE.
It enforces hard chart rules: no in-figure titles, captions below the figure, analysis text right after each one, max 3 figures of the same type, print-ready font sizes, a single 300 DPI PNG output, and 7 journal color palettes like Nature, Science, Cell and IEEE.How are flowcharts and technical diagrams generated?For non-data diagrams such as technical roadmaps and model structure charts, it bundles drawio-skill v2.1.0, which auto-generates them and exports to PDF/PNG.
For non-data diagrams such as technical roadmaps and model structure charts, it bundles drawio-skill v2.1.0, which auto-generates them and exports to PDF/PNG.What happens during final acceptance checks?Before the paper is considered done it runs a text gate, real document compilation, page-spec checks, figure/table numbering checks, bidirectional citation verification, and a five-way consistency check between the AI usage log, the AI declaration and the AI usage report.
Before the paper is considered done it runs a text gate, real document compilation, page-spec checks, figure/table numbering checks, bidirectional citation verification, and a five-way consistency check between the AI usage log, the AI declaration and the AI usage report.What are similar projects?codex-claude-academic-skills, modex-mh-agent, PaperSpine, FigGenie-paper-diagram-skill, bzd-math-modeling-skills, cumcm-paper-hand-skill, SenseNova-Skills, claude-scholar
codex-claude-academic-skills, modex-mh-agent, PaperSpine, FigGenie-paper-diagram-skill, bzd-math-modeling-skills, cumcm-paper-hand-skill, SenseNova-Skills, claude-scholarWhich topics is it in?paper-writing, academic-writing, diagrams, excel, latex, students, mathematics, document-processing, claude-skills, workflow, data-visualization, pdf
paper-writing, academic-writing, diagrams, excel, latex, students, mathematics, document-processing, claude-skills, workflow, data-visualization, pdf