brycewang-stanford/stat-writing

<!-- ╔══════════════════════════════════════════════════════════════╗ ║ 本文件为开源 Skill 原始文档，收录仅供学习与研究参考 ║ ║ CoPaper.AI 收集整理 | https://copaper.ai ║ ╚══════════════════════════════════════════════════════════════╝ 来源仓库: https://github.com/fuhaoda/stats-paper-writing-agent-skills 项目名称: stats-paper-writing-agent-skills 开源协议: MIT License 收录日期: 2026-04-02 声明: 本文件版权归原作者所有。此处收录旨在为社会科学实证研究者 提供 AI Agent Skills 的集中参考。如有侵权，请联系删除。 -->

Statistical Writing (One Skill)

This is a single "workbench" skill for statistical manuscripts written in LaTeX.

Guidance is split into focused references under references/. Deterministic checks live in scripts/. Ready-to-use templates live in assets/.

Positioning: Hybrid + JDS profile

Default behavior is journal-agnostic. For Journal of Data Science (JDS), apply the JDS profile:

• Maintain strong literature positioning and explicit novelty.
• Require clean cross-referencing and cleaned BibTeX.
• Prefer vector graphics for figures in the manuscript.
• Enable line numbers for review drafts.
• Include reproducibility artifacts (code/data/supplement) when possible.

When to use this skill

Use this skill when the user wants to:

1. Generate compliant front matter (title, abstract, keywords).
2. Expand outlines into complete sections in LaTeX.
3. Audit a manuscript for structure, style, references, and reproducibility quality.
4. Draft reviewer reports.
5. Draft point-by-point response letters.
6. Scaffold a book manuscript from a chapter plan.

Inputs to ask for (minimal)

Prefer file paths over pasted text.

• Manuscript tasks: root TeX file (for example main.tex).
• Reference checks: BibTeX file(s) (for example refs.bib).
• Response letters: full reviewer/editor comments + revised text if available.
• Reviewer report: manuscript or extended abstract being reviewed.
• Book manuscript: chapter list, audience, tone, and desired notation style.

If details are missing, proceed with placeholders like odo{...} and ask only critical questions.

Output conventions

Unless the user requests otherwise:

• Return LaTeX-ready output.
• For audits: rank issues as HIGH/MED/LOW with concrete fixes.
• For rewrites: include revised text and short change log.
• Never invent results or citations. Use odo{add citation} or odo{verify result}.

Task routing map

Open only the reference files needed for the task.

• Title: references/10-title.md
• Abstract: references/11-abstract.md
• Keywords: references/12-keywords.md
• Outline to section drafting: references/60-outline-to-section.md
• Introduction: references/20-introduction.md
• Data: references/21-data.md
• Methods: references/22-methods.md
• Simulation (ADEMP): references/23-simulation.md
• Application/Results: references/24-application.md
• Discussion: references/25-discussion.md
• Other sections: references/30-other-sections.md
• General style/storyline: references/31-general-style.md
• English pitfalls: references/32-english.md
• BibTeX/natbib: references/40-bibtex-natbib.md
• Labels/cross-references: references/41-cross-referencing.md
• Reviewer report: references/50-review-report.md
• Response to reviewers: references/51-response-to-reviewers.md
• Research proposal/project cycle: references/70-project-proposal.md
• Book manuscript workflow: references/71-book-manuscript.md
• Tooling/reproducibility: references/80-tooling.md

Built-in assets

• Response letter (LaTeX): assets/response-letter-template.tex
• Reviewer report (Markdown): assets/reviewer-report-template.md
• Reviewer report (LaTeX): assets/reviewer-report-template.tex
• Generic section skeleton: assets/section-skeleton.tex
• Manuscript starter (LaTeX): assets/manuscript-template.tex
• Book manuscript starter (LaTeX): assets/book-manuscript-template.tex

Optional scripts (deterministic checks)

• Manuscript checks: python scripts/check_tex.py path/to/main.tex
• Citation/BibTeX checks: python scripts/check_bib.py --tex path/to/main.tex --bib path/to/refs.bib
• Combined run: python scripts/audit_paper.py --tex path/to/main.tex --bib path/to/refs.bib

These checks are heuristic and do not compile LaTeX.

Workflows

Workflow A - Finished paper to abstract + keywords

1. Read introduction/methods/results/discussion.
2. Use references/11-abstract.md and references/12-keywords.md.
3. Draft abstract (default 6-8 sentences, acceptable 4-10, no citations, no math notation).
4. Draft 6-10 keywords, alphabetized, avoid repeating title terms.
5. Return:
   • egin{abstract}...nd{abstract}
   • \keywords{...} (or venue-specific command)
   • short compliance checklist.

Workflow B - Manuscript audit

1. Run check_tex.py (and check_bib.py if .bib exists).
2. Use references/31-general-style.md, references/40-bibtex-natbib.md, and section-specific references.
3. Return top issues ranked by severity and concrete LaTeX edits.
4. For JDS profile, explicitly call out line numbers, vector graphics, cleaned BibTeX, and reproducibility supplement readiness.

Workflow C - Reviewer report drafting

1. Use references/50-review-report.md.
2. Write summary + overall assessment + numbered major/minor comments.
3. Keep tone constructive and professional.
4. If requested, output using assets/reviewer-report-template.tex.

Workflow D - Response to reviewers

1. Use references/51-response-to-reviewers.md.
2. Structure by Editor, Associate Editor, Reviewer sections.
3. For every comment: quote, respond, quote manuscript change, add location.
4. If requested, render with assets/response-letter-template.tex.

Workflow E - Outline to full section

1. Identify section type.
2. Use references/60-outline-to-section.md + relevant section reference.
3. Expand bullets into coherent paragraphs with transitions.
4. Use placeholders where information is missing.

Workflow F - Book manuscript scaffolding

1. Use references/71-book-manuscript.md.
2. Start from assets/book-manuscript-template.tex.
3. Build frontmatter/mainmatter/backmatter and chapter map.
4. Keep notation generic by default; add optional custom notation block only when requested.

Copy/paste prompt patterns

Abstract + keywords

"Use stat-writing. Read main.tex and draft a compliant abstract (default 6-8 sentences, acceptable 4-10; no citations; no math notation) and 6-10 alphabetical keywords. Output LaTeX blocks."

Full audit

"Use stat-writing. Audit main.tex (+ refs.bib). Run scripts if allowed. Return top issues with HIGH/MED/LOW and patch-ready LaTeX fixes."

Response to reviewers

"Use stat-writing. Here are reviewer/editor comments. Write a point-by-point response letter with quoted manuscript revisions and locations."

Reviewer report

"Use stat-writing. Draft a reviewer report with summary, overall assessment, numbered major comments, and numbered minor comments."

Book manuscript

"Use stat-writing. Start a book manuscript from chapter bullets using the book template. Produce frontmatter/mainmatter/backmatter and chapter-by-chapter drafting plan."