Paper Writer Role (Math Modeling Paper Generator)

What it does

• Takes the contest problem, modeling deliverables, and tables/figures produced by actually executed code, and produces a complete Word paper (完整论文.docx); on explicit request it also builds a LaTeX source project and a genuinely compiled PDF with identical content.
• Enforces an entry gate: without the problem, both modeling deliverables, runnable code, real result tables, candidate figures (≥3 per category covering every sub-question), and verified references, it refuses to write and sends work back upstream.
• Builds an internal Claim–Evidence map per sub-question (claim, formula, result-table location, at least one formal figure, code output or verified citation) and forbids fabricating evidence.
• Treats the current-year official contest rules as top priority (page limits, margins, abstract page, submission format); bundled templates are only a fallback baseline, never a substitute.
• Requires dual-engine literature search plus opening the DOI/publisher page to verify metadata — copying an upstream reference list does not count.
• Finishes with deterministic checks: paper_format.py validate, office/validate.py, equations.py verify-conversion, and for LaTeX latex_paper.py doctor/build/validate (image DPI, body/appendix page boundaries, font embedding, hash binding). All exit codes must be 0, and independent W1/W2 QA subagents must return PASS before delivery.

Who it's for

• Students and advisors writing CUMCM / MCM-ICM style math modeling papers.
• Anyone who must keep a Word and a LaTeX version perfectly consistent under strict formatting rules.
• Users who want AI drafts grounded strictly in real computed results instead of plausible-sounding filler.

Examples

1. Word only: hand over modeling and coding outputs; the skill assembles 完整论文.docx in the official structure and verifies it with paper_format.py validate --contest cumcm --rendered-pages N.
2. Word + LaTeX: say "I also need LaTeX"; it runs latex_paper.py init to copy the official template, compiles with xelatex, then runs validate --min-image-dpi 300 --questions q1 q2 q3.
3. Missing evidence: if question 2 has no result table, writing halts and the task returns to the modeling/coding roles before the W1 gate is re-run.

Read the instructions in this file and follow them to help me:
https://raw.githubusercontent.com/XiaoMaColtAI/math-modeling-skill/HEAD/dsh-plugin/math-modeling-agent/skills/math-modeling/references/roles/论文手/SKILL.md

What I want: (describe your task here)

Install the skill I found on Claude Skill Mart.
Copy the dsh-plugin/math-modeling-agent/skills/math-modeling/references/roles/论文手 folder from the GitHub repo XiaoMaColtAI/math-modeling-skill into my ~/.claude/skills/xiaomacoltai-skill-2/.
When it's done, tell me in one line what this skill can do.