download-analyze-paper

name: download-analyze-paper description: Downloads an arXiv paper given its ID/URL, creates a target directory, extracts the LaTeX source, and generates a reviewer-level analytical note. Use when the user wants to download a paper and generate notes for it.

Paper Download and Analysis Workflow

This skill provides a self-contained workflow for downloading arXiv papers and performing a deep reviewer-level analysis.

Instructions

When the user asks to download an arXiv paper and generate a note, follow these steps.
By default, you should assume a BASE_DIR under which you always create two subfolders:

• BASE_DIR/papers 用于存放论文 LaTeX 源码（每篇一个子目录）
• BASE_DIR/notes 用于存放对应的 Markdown 审稿笔记

约定规则：

• 若用户显式给出了一个目录（例如 reviewer），则将该目录视为 BASE_DIR。
• 若用户未给出目录，只是说“下载论文并生成笔记”，则将仓库根目录视为 BASE_DIR。

1. Extract Information:

   • Extract the ARXIV_ID from the user's input (e.g., from https://arxiv.org/abs/2602.11124, the ID is 2602.11124).
   • Determine the METHOD_NAME (e.g., "PhyCritic"). If the user doesn't provide one, try to extract it from the paper title or ask the user.
   • Determine the BASE_DIR:
     • If the user mentions a directory like reviewer ("下载到reviewer并生成笔记"), then BASE_DIR = reviewer.
     • Otherwise, use the workspace root as BASE_DIR.
   • Determine the TARGET_DIR where the paper should be saved: TARGET_DIR = <BASE_DIR>/papers/<METHOD_NAME> (e.g., reviewer/papers/PhyCritic).
   • Determine the TARGET_NOTES_DIR where the output markdown note should be saved: TARGET_NOTES_DIR = <BASE_DIR>/notes/<METHOD_NAME>.md (e.g., reviewer/notes/PhyCritic.md).

2. Locate Skill Assets:

   • Find the path to this skill's directory. In this project it is typically .cursor/skills/paper-reading-skill（如已全局安装则可能为 ~/.cursor/skills/download-analyze-paper，需按实际路径调整）。
   • You will need the script scripts/download_arxiv.sh and the prompt file reviewer_prompt.md located inside the skill directory.

3. Download and Extract Paper Source:

   • Execute the generalized bash script provided by this skill to download the paper. Make sure to use the absolute path to the skill folder.

     bash <SKILL_DIR>/scripts/download_arxiv.sh <ARXIV_ID> <TARGET_DIR>
     

4. Generate Analytical Note:

   • Use the Read tool to read the contents of <SKILL_DIR>/reviewer_prompt.md.
   • Use the Task tool (with subagent_type: generalPurpose) to analyze the downloaded paper.
   • In the Task prompt, include the full text of reviewer_prompt.md so the subagent knows exactly how to analyze the paper.
   • Instruct the subagent to read the LaTeX source in <TARGET_DIR>.
   • Instruct the subagent to strictly follow the criteria in the prompt.
   • Instruct the subagent to save the output markdown note to <TARGET_NOTES_DIR> (e.g., reviewer/notes/PhyCritic.md).
   • Tell the subagent to use $$ for block math formulas and $ for inline math.
   • Important: Markdown does not support \begin{aligned}, \begin{array}, \begin{cases}, or other LaTeX environments. Only use simple inline ($...$) or block ($$...$$) math syntax.
   • Tell the subagent to perform web searches for section 4 of the prompt ("与当前领域主流共识及反对观点的关系").

5. Confirm Completion:

   • Once the subagent finishes, report success to the user and mention the path to the newly created markdown note.

Example Usage

User: "下载论文 https://arxiv.org/abs/2602.11124 到某个目录（例如 reviewer），并生成审稿笔记。"

Agent Action:

1. 解析得到例如 ARXIV_ID = 2602.11124，METHOD_NAME = <METHOD_NAME>，BASE_DIR = <BASE_DIR>，因此：
   • TARGET_DIR = <BASE_DIR>/papers/<METHOD_NAME>
   • TARGET_NOTES_DIR = <BASE_DIR>/notes/<METHOD_NAME>.md
2. Run bash <SKILL_DIR>/scripts/download_arxiv.sh 2602.11124 <BASE_DIR>/papers/<METHOD_NAME>.
3. Read <SKILL_DIR>/reviewer_prompt.md.
4. Run Task tool prompting the subagent to read <BASE_DIR>/papers/<METHOD_NAME>/main.tex, follow the reviewer prompt, and write the result to <BASE_DIR>/notes/<METHOD_NAME>.md.