词映 AI导航
工具库Skills 技能
我的收藏
词映 AI导航

Sanmu Paper Reader for Claude Code and Codex: Trace Evidence in Research Papers

Use Sanmu Paper Reader to turn a paper, supplementary information, or a local Markdown extract into notes that point back to source sections. It records units and test conditions alongside values and identifies gaps before comparing results from different studies.

中文

Install in Claude Code、Codex

Use Claude Code or Codex with access to local files. A Markdown extract or text-based PDF is enough for text reading. Standalone helper scripts use Python 3.10 or later. Figure tools additionally require PyMuPDF and Pillow. The complete plugin requires Node.js 22.20 or later and Python 3.10 or later.

  1. Get the pinned GitHub version and keep the complete `skills/sanmu-paper-reader/` directory, including `SKILL.md`, `references/`, and `scripts/`.
  2. Copy the directory into `~/.claude/skills/` for Claude Code or `~/.agents/skills/` for Codex. Open a fresh session with a sample Markdown document or text-based PDF.
  3. Use the skill’s `requirements.txt` in a virtual environment when you need figure cropping. Follow the repository README for the complete plugin if you need the browser workspace or the MCP server.

Example task

Save the two lines below in `paper-demo.md` in a fresh project, then give the final instruction to your agent.

Section 1, Methods: Sensor A records ten samples per minute at 20 °C.
Section 2, Results: A averages 2 mW. A separate experiment abstract gives 1.5 mW for B and omits its sampling rate.

Use sanmu-paper-reader to trace each number and identify the conditions given in this fictional document. Can we conclude that B uses less power? Reply in chat. Use the text only.

What to check

The notes should point to Sections 1 and 2 for A, identify the abstract and missing test conditions for B, and explain why the supplied material cannot establish which sensor uses less power.

Original SKILL.md on GitHub · Updated

Frequently asked question

Do I need Zotero if I already have a Markdown extract?

You can begin with a local Markdown extract. Zotero access is only needed for tasks that select files or metadata from a personal library. Cite section titles when the extract has no reliable page mapping.