← 返回 Skills 市场
myjackcat

Auto Paper Writer

作者 MyJackCat · GitHub ↗ · v1.0.0 · MIT-0
cross-platform ⚠ suspicious
128
总下载
0
收藏
0
当前安装
1
版本数
在 OpenClaw 中安装
/install auto-paper-writer
功能描述
自动撰写科研论文的完整 AI pipeline。当用户说"撰写[方向]论文"、"自动写论文"、"生成论文"时触发。功能包括:(1) 搜索 ArXiv 最新论文并下载原始 PDF;(2) 总结现有工作局限性;(3) 设计创新模型框架;(4) 撰写双栏 LaTeX 论文(IEEEtran格式);(5) 生成科研级 m...
使用说明 (SKILL.md)

Auto Paper Writer

自动撰写科研论文的完整 pipeline。从文献调研到 PDF 输出的端到端自动化。

工作流程

Step 1: 搜索最新论文

使用 ArXiv API 搜索相关方向的最新论文:

# PowerShell 调用 ArXiv API
$query = "your search terms"
$url = "https://export.arxiv.org/api/query?search_query=all:$query&start=0&max_results=10&sortBy=submittedDate&sortOrder=descending"

优先搜索 ICLR/NeurIPS/AAAI/ICML 等顶会论文,以及与用户方向相关的高引用工作。

Step 1.5: 保存搜索论文

将搜索到的论文信息保存到文献文件夹:

# 创建文献文件夹
$ref_dir = "$paper_dir\references"
New-Item -ItemType Directory -Path $ref_dir -Force

# 保存论文列表(Markdown 格式)
$ref_content = @"
# 文献调研

## 搜索时间:$(Get-Date -Format "yyyy-MM-dd HH:mm")
## 搜索方向:$direction
## 搜索关键词:$keywords

---

### 论文列表

| # | 论文标题 | 作者 | 日期 | arXiv ID |
|---|---------|------|------|----------|
"@

foreach ($entry in $xml.feed.entry) {
    $title = $entry.title -replace '\s+', ' '
    $authors = ($entry.author | ForEach-Object { $_.name }) -join ', '
    $published = $entry.published.Substring(0,10)
    $arxiv_id = $entry.id.Split('/')[-1]
    $ref_content += "`n| - | $title | $authors | $published | [$arxiv_id](https://arxiv.org/abs/$arxiv_id) |"
}

$ref_content += "@"
$ref_content | Out-File -FilePath "$ref_dir\paper_references.md" -Encoding UTF8

保存内容:

  • references/paper_references.md — 论文列表(标题/作者/日期/链接)
  • references/*.pdf — 原始论文 PDF 文件

Step 1.5: 下载论文原文 PDF

从 ArXiv 下载每篇论文的原始 PDF,文件用论文标题命名

# ArXiv PDF 下载地址格式
# https://arxiv.org/pdf/{arxiv_id}.pdf

foreach ($entry in $xml.feed.entry) {
    $arxiv_id = $entry.id.Split('/')[-1] -replace 'v\d+$', ''
    $title = $entry.title -replace '\s+', ' ' -replace '[^\w\s-]', '' -replace '\s', '_'
    $pdf_url = "https://arxiv.org/pdf/$arxiv_id.pdf"
    $pdf_path = "$ref_dir\$title.pdf"

    # 截断过长文件名,保留前 50 字符
    if ($title.Length -gt 50) {
        $title = $title.Substring(0, 50)
        $pdf_path = "$ref_dir\$title.pdf"
    }

    try {
        Invoke-WebRequest -Uri $pdf_url -OutFile $pdf_path -UseBasicParsing
        Write-Host "Downloaded: $title.pdf"
    } catch {
        Write-Host "Failed to download: $arxiv_id"
    }
}

文件命名规则

  • 用论文标题命名,不是 arXiv ID
  • 空格替换为下划线
  • 特殊字符清理
  • 过长截断前 50 字符

下载结果示例:

  • references/Bidirectional_Cross-Modal_Prompting_for_Event-Frame.pdf
  • references/MM-WebAgent_A_Hierarchical_Multimodal_Web_Agent.pdf
  • ...

重要:此步骤是标准工作流程的必要环节,必须执行!

Step 2: 分析局限性

阅读摘要和核心方法,总结现有工作的局限性

  • 计算开销大 / 参数量过大
  • 需要海量标注数据
  • 模态对齐/融合不完善
  • 泛化能力有限
  • 推理速度慢

Step 3: 设计创新模型框架

  • 起一个简洁有力的模型名字
  • 提出 3-5 个核心创新点
  • 画整体架构框图(模块化设计)

Step 4: 撰写论文

按标准科研论文结构,使用 IEEEtran 双栏格式

\documentclass[conference]{IEEEtran}

论文结构:

  1. Abstract - 简明扼要,150-250 词
  2. Introduction - 研究背景、动机、贡献(列出 3-5 条贡献)
  3. Related Work - 梳理相关工作,对比分析
  4. Proposed Method - 详细介绍方法
  5. Experiments - 实验设置、基线对比
  6. Conclusion - 总结与未来工作
  7. References - 格式规范的参考文献

使用标准 LaTeX 公式、表格、\ref{} 交叉引用。

Step 5: 生成科研级配图

使用 matplotlib 生成 300dpi 高质量图片:

import matplotlib.pyplot as plt
plt.rcParams['font.size'] = 10
plt.rcParams['figure.dpi'] = 300

# 图1: 模型框架图 - 整体架构,模块化,箭头标注
# 图2: 方法细节图 - 核心创新机制
# 图3: 实验结果图 - 对比曲线、柱状图

Step 6: 编译 PDF

使用 TeX Live(不是 Tectonic)编译:

$texlive = "D:	exlive\2026\bin\windows\pdflatex.exe"
$paper_dir = "C:\Users\29064\Desktop\PaperName"

# 必须运行两次!第一次生成 .aux,第二次解析交叉引用
& $texlive "$paper_dir\paper.tex" -interaction=nonstopmode -output-directory="$paper_dir"
& $texlive "$paper_dir\paper.tex" -interaction=nonstopmode -output-directory="$paper_dir"

关键规则

  • 编译失败时绝不简化 .tex 内容,保留完整让用户在 Overleaf 自行处理
  • 编译时必须在 tex 文件所在目录运行(否则图片相对路径找不到)
  • 不要用 Tectonic(与 IEEEtran 不兼容)

Step 7: 保存到桌面

输出路径:C:\Users\29064\Desktop\[论文名字]\ 包含文件:

  • paper.tex - 主论文文件
  • paper.pdf - 编译后的 PDF
  • figure1.png / figure2.png / figure3.png - 配图
  • references/paper_references.md - 文献调研列表
  • references/*.pdf - 原始论文 PDF

Step 8: 清理临时文件

论文撰写完成后,删除所有临时脚本文件

  • 搜索脚本(search_*.ps1
  • 下载脚本(download_*.ps1
  • 配图生成脚本(generate_*.py
  • 其他中间文件
Remove-Item "$env:USERPROFILE\Desktop\*.ps1" -Force
Remove-Item "$env:USERPROFILE\Desktop\*.py" -Force
Remove-Item "$env:USERPROFILE\Desktop\*_v*.pptx" -Force

核心原则

  1. 生成后立即编译:.tex 完成后立刻用 TeX Live 编译
  2. 保留完整性:不简化内容,编译失败时输出完整 .tex
  3. 双栏 IEEEtran:标准学术论文格式
  4. 科研级配图:300dpi,清晰标注

参考文件

安全使用建议
Before installing or running this skill, consider the following: (1) It is Windows/PowerShell–centric and hard-codes user-specific paths (C:\Users\29064\Desktop and D:\texlive\2026\...), so it will likely fail or act unexpectedly on other systems. (2) The SKILL.md contains overbroad deletion commands that remove all .ps1 and .py files from the Desktop — this can delete unrelated or important scripts. Remove or limit these Remove-Item lines before use. (3) Filenames use sanitized paper titles rather than stable arXiv IDs, increasing risk of collisions/overwrites and filesystem issues; prefer arXiv IDs for downloaded PDFs. (4) The skill assumes pdflatex exists at a specific path; verify and adjust to your TeX installation or run the compilation step manually. (5) Because the skill will download many PDFs and write files to your Desktop, run it in a sandbox or dedicated folder and back up important files first. (6) Review the generated content for plagiarism and experimental claims before submission — the author already warns it does not guarantee originality or experimental truth. If the SKILL.md is edited to remove destructive file deletion, remove hard-coded paths, and make platform assumptions explicit (or provide configurable paths), my assessment would move toward benign.
功能分析
Type: OpenClaw Skill Name: auto-paper-writer Version: 1.0.0 The skill performs high-risk operations including executing PowerShell commands to download external files and running local binaries for LaTeX compilation. A significant concern is found in the cleanup logic of SKILL.md (Step 8), which uses broad wildcards to delete all .ps1 and .py files from the user's desktop, potentially causing unintended data loss. Additionally, the use of hardcoded absolute file paths (e.g., C:\Users\29064\Desktop) suggests the code was tailored for a specific environment or lacks proper input sanitization for general use.
能力评估
Purpose & Capability
The skill claims an end-to-end pipeline for writing papers which reasonably includes searching arXiv, summarizing, producing LaTeX and figures, and compiling PDFs. However the instructions assume a Windows/PowerShell environment, hard-code user-specific paths (C:\Users\29064\Desktop and D:\texlive\2026\bin\windows\pdflatex.exe) and insist on behaviors (naming PDFs by title rather than stable arXiv IDs, mandatory immediate compilation) that are incongruent with a portable, general-purpose 'auto-paper-writer' skill.
Instruction Scope
Most steps are within the stated purpose (calling arXiv API, downloading PDFs, generating figures, writing LaTeX). But the SKILL.md also instructs deleting broad classes of files from the user's Desktop (Remove-Item "$env:USERPROFILE\Desktop\*.ps1" and similar), uses file-naming and file-placement policies that could overwrite or move unrelated user files, and enforces 'this step must execute' language that gives the agent strong, potentially destructive autonomy.
Install Mechanism
This is an instruction-only skill (no install spec), which reduces risk of writing/executing arbitrary new code. However it assumes external binaries (PowerShell environment and a TeX Live binary at a hard-coded path) that are not declared — this mismatch between assumptions and declared requirements is noteworthy.
Credentials
The skill requests no credentials or env vars, which is appropriate. But it directly references and manipulates user filesystem locations ($env:USERPROFILE\Desktop, hard-coded D: and C: paths) and issues deletion commands that can remove arbitrary .ps1/.py files on the Desktop. Access to filesystem paths in itself can be necessary, but the broad deletion behavior is disproportionate to the stated purpose and risky.
Persistence & Privilege
The skill is not always-enabled, and does not request persistent privileges or modification of other skills. Autonomous invocation is allowed (platform default), which combined with the above risks could increase impact, but on its own there is no unusual persistence flag.
如何使用
  1. 确保已安装 OpenClaw(本地或 Docker 部署)
  2. 在对话框中输入安装命令:/install auto-paper-writer
  3. 安装完成后,直接呼叫该 Skill 的名称或使用 /auto-paper-writer 触发
  4. 根据 Skill 的参数说明提供必要输入,即可获得结构化输出
版本历史
v1.0.0
Initial release: 自动撰写科研论文完整pipeline,支持从ArXiv搜索到PDF编译的端到端自动化
元数据
Slug auto-paper-writer
版本 1.0.0
许可证 MIT-0
累计安装 0
当前安装数 0
历史版本数 1
常见问题

Auto Paper Writer 是什么?

自动撰写科研论文的完整 AI pipeline。当用户说"撰写[方向]论文"、"自动写论文"、"生成论文"时触发。功能包括:(1) 搜索 ArXiv 最新论文并下载原始 PDF;(2) 总结现有工作局限性;(3) 设计创新模型框架;(4) 撰写双栏 LaTeX 论文(IEEEtran格式);(5) 生成科研级 m... 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件,目前累计下载 128 次。

如何安装 Auto Paper Writer?

在 OpenClaw 或 Claude Code 对话框中运行命令「/install auto-paper-writer」即可一键安装,无需额外配置。

Auto Paper Writer 是免费的吗?

是的,Auto Paper Writer 完全免费,采用 MIT-0 许可证,可自由下载、安装和使用。

Auto Paper Writer 支持哪些平台?

Auto Paper Writer 跨平台运行,可在任意部署了 OpenClaw / Claude Code 的环境中使用(cross-platform)。

谁开发了 Auto Paper Writer?

由 MyJackCat(@myjackcat)开发并维护,当前版本 v1.0.0。

💬 留言讨论