What is this?
A collection of Skills for the Codex and Claude Code coding agents that guides you through the whole math-modeling competition workflow — from understanding the problem statement to a judge-style review of your finished paper.
A collection of Skills for the Codex and Claude Code coding agents that guides you through the whole math-modeling competition workflow — from understanding the problem statement to a judge-style review of your finished paper.Is this a separate app, or something that lives inside my coding agent?It lives inside your coding agent: each piece is a Skill folder you copy into Codex or Claude Code, and the agent then runs the checks, audits and reviews on your paper. There is no app, server or build step of its own.
It lives inside your coding agent: each piece is a Skill folder you copy into Codex or Claude Code, and the agent then runs the checks, audits and reviews on your paper. There is no app, server or build step of its own.Do I need Claude Code specifically, or does Codex work too?Either: the collection targets Codex and Claude Code, with extra Claude Code configuration in integrations/claude-code/. No Python setup, dependencies or backend service are needed beyond the agent itself.
Either: the collection targets Codex and Claude Code, with extra Claude Code configuration in integrations/claude-code/. No Python setup, dependencies or backend service are needed beyond the agent itself.How much should I trust the scores and rank predictions?As reference only: it is explicitly not an official competition tool, and highly original solutions outside conventional grading paths may be undervalued — an experienced advisor or human judge should still make the final call.
As reference only: it is explicitly not an official competition tool, and highly original solutions outside conventional grading paths may be undervalued — an experienced advisor or human judge should still make the final call.Is this built for the Chinese national contest (CUMCM) only, or other modeling competitions too?It is built around China's CUMCM national contest (高教社杯), with rules distilled from its 2020-2025 editions. Rules and AI policies differ between contests and change yearly, so verify your own competition's current requirements.
It is built around China's CUMCM national contest (高教社杯), with rules distilled from its 2020-2025 editions. Rules and AI policies differ between contests and change yearly, so verify your own competition's current requirements.Does my operating system matter?No — the skills run wherever Codex or Claude Code run: Windows, macOS or Linux. They are plain folders the agent reads, with nothing OS-specific inside.
No — the skills run wherever Codex or Claude Code run: Windows, macOS or Linux. They are plain folders the agent reads, with nothing OS-specific inside.My contest problems and paper are not public yet — anything I should be careful about?Yes: the README warns users never to submit unpublished contest problems, unauthorized papers, personal identity information, API keys, tokens or internal materials to the public repository.
Yes: the README warns users never to submit unpublished contest problems, unauthorized papers, personal identity information, API keys, tokens or internal materials to the public repository.If I'm not sure which skill I need, where do I start?Start with the master skill bzd-modeling-workflow: it detects your current competition stage, picks the specialized skills to run, records stage artifacts and suggests next steps instead of mechanically running every check.
Start with the master skill bzd-modeling-workflow: it detects your current competition stage, picks the specialized skills to run, records stage artifacts and suggests next steps instead of mechanically running every check.What does the final review of my paper actually return?The bzd-review-paper skill reviews your paper like a competition judge: it derives a scoring rubric for your specific problem from the 2020-2025 official CUMCM grading criteria, then returns a percent quality score, a format quality coefficient, a rank estimate calibrated to your competition environment, itemized deductions and an HTML report.
The bzd-review-paper skill reviews your paper like a competition judge: it derives a scoring rubric for your specific problem from the 2020-2025 official CUMCM grading criteria, then returns a percent quality score, a format quality coefficient, a rank estimate calibrated to your competition environment, itemized deductions and an HTML report.Can it flag writing that looks AI-generated?Yes — bzd-paper-aigc-auditor audits the language for AI traces, templated modeling, model stacking and fake improvements, returning a risk range with per-section evidence and suggestions for making the text sound more human.
Yes — bzd-paper-aigc-auditor audits the language for AI traces, templated modeling, model stacking and fake improvements, returning a risk range with per-section evidence and suggestions for making the text sound more human.The contest expects AI usage to be disclosed — can it help with that?Yes. bzd-ai-usage-disclosure generates or checks your AI-tools usage statement and detail materials against your real usage records, verifying completeness, consistency, anonymity and responsibility boundaries.
Yes. bzd-ai-usage-disclosure generates or checks your AI-tools usage statement and detail materials against your real usage records, verifying completeness, consistency, anonymity and responsibility boundaries.What are similar projects?cumcm-step-review, gathered-scenes-zine-skill, anti-defensive-writing-Skill, modex-mh-agent, PaperSpine, codex-claude-academic-skills, Auto-claude-code-research-in-sleep, nature-skills
cumcm-step-review, gathered-scenes-zine-skill, anti-defensive-writing-Skill, modex-mh-agent, PaperSpine, codex-claude-academic-skills, Auto-claude-code-research-in-sleep, nature-skillsWhich topics is it in?paper-writing, aigc, academic-writing, math, students, academic, template, ai-tools, skill, html, agent-skills, codex
paper-writing, aigc, academic-writing, math, students, academic, template, ai-tools, skill, html, agent-skills, codex