research-paper-writing

これはレビュー時に保存された抜粋です。完全で最新の内容は外部ソースを確認してください。

---
name: research-paper-writing
description: Use when drafting, reviewing, rewriting, or polishing academic research papers, workshop/journal submissions, manuscript sections, claims, citations, methodology, results, figures, tables, threats to validity, anonymized review versions, or camera-ready text. Do not use for general product docs, blog posts, or code comments.
---

# Research Paper Writing Skill

Use this skill when drafting, reviewing, rewriting, or polishing a research paper for a conference, workshop, journal, or internal academic review.

## Goal

Write a standalone scientific manuscript that reads like work produced by domain researchers for peer review. The paper must communicate:

- the problem,
- the gap in prior work,
- the method,
- the evidence,
- the results,
- the limitations,
- and the contribution.

The manuscript must not read like a report about how the document was generated.

## Core principles

- State claims precisely.
- Separate measured results from interpretation.
- Keep scope boundaries clear without sounding defensive.
- Cite external ideas, facts, definitions, methods, and prior results.
- Do not cite sources that do not support the sentence.
- Do not mention the target venue, writing tools, AI tools, build process, plotting packages, or internal files in the paper body.
- Keep source/build details in comments, scripts, README files, or artifact documentation, not in the manuscript text.

## Source hierarchy

Prefer sources in this order:

1. The authors' own data, experiments, code, and measurements for reported results.
2. Official documentation, standards, specifications, white papers, technical reports, audits, or repositories for system-specific claims.
3. Peer-reviewed papers and strong preprints for theory, related work, risks, methods, and empirical findings.
4. Official tool or infrastructure documentation only when the tool is part of the research method.
5. Blogs, dashboards, tutorials, and news only for context, never for core scientific claims unless no stronger source exists.

## What belongs in the paper body

Include:

- problem motivation and research gap,
- research questions or hypotheses,
- related work and how this paper differs,
- background needed to understand the contribution,
- experimental setup or method,
- variables, assumptions, scenarios, and metrics,
- tools that are part of the scientific method,
- tables and figures showing research content,
- results scoped to the evaluated setting,
- interpretation supported by data and citations,
- threats to validity,
- conclusion and future work.

## What must not appear in the paper body

Do not include document-production details, such as:

- target conference or venue name,
- statements justifying the topic through a venue call for papers,
- typesetting system or template decisions,
- plotting, drawing, or diagramming package names,
- "generated from CSV", "rendered by native figures", or similar build-process wording,
- internal file names or paths,
- build commands,
- script names used only to produce the manuscript,
- AI assistants, prompts, generated-text workflow, or editing process,
- citations to document-production tools.

Exception: mention a tool only when it is part of the scientific experiment itself, not merely part of writing, formatting, plotting, or packaging the paper.

## Venue and review mode

Use venue requirements only to configure the manuscript.

Allowed outside the paper body:

- template choice,
- page limits,
- anonymization settings,
- citation style,
- source comments,
- README notes,
- submission checklist.

Forbidden in the paper body:

- "This paper is written for ..."
- "This fits the conference because ..."
- citations to the venue call for papers,
- explanations of the template or formatting tools.

For double-blind review:

- anonymize authors and affiliations,
- remove acknowledgements that identify the authors,
- remove identifying artifact links if required,
- cite the authors'