菜单
Extract text from papers, summarize them, and create reading notes. Use when reading or summarizing a paper, and automatically after downloading one.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Build the manuscript to PDF/DOCX/HTML, create figures and diagrams, manage cross-references, and run writing-quality checks (lint, grammar, readability).
Search academic literature, manage the search queue, download papers, and verify sources. Use when the user wants to find papers, when working search-queue.md, or when adding references.
Pre-submission checks, bibliography verification, and peer-review responses. Use when the manuscript nears submission, when verifying references, or when responding to reviewers.
Create writing plans, section drafts, and manuscript prose. Use when drafting or editing any part of the paper, building evidence chains from notes, or planning the writing.
Conventions for projects that generate their own results — running models, labeling data, computing statistics. Use when writing or running experiment code, or when drafting sections that report original results.
Plain-language Git for a non-technical user — snapshots, backups, undo, and disaster recovery. Use for any Git operation beyond a routine commit, and whenever the user says save, back up, undo, or go back.
| name | reading-notes |
| description | Extract text from papers, summarize them, and create reading notes. Use when reading or summarizing a paper, and automatically after downloading one. |
Reading notes are objective (what does the paper say?) — create them automatically, no approval needed. One note per paper in notes/.
make extract-text PDF="sources/filename.pdf" to get the textYYYY-MM-DD-authorlastname-year---short-title.md (e.g., 2026-03-08-smith-2024---cognitive-offloading.md)---
title: "Author Year - Short Title"
date: YYYY-MM-DD
tags: [relevant, topic, tags]
doi: "10.xxxx/..."
---
templates/note.md: Summary, Key Findings, Methodology, Relevance to This Work, Questions & Follow-ups, plus one key quote the user can paste into the manuscriptnotes/."make extract-text PDF="sources/paper.pdf" — extract one PDFmake extract-all — extract every PDF in sources/make new-note TITLE="Paper Title" — create a note from the templatemake search-notes QUERY="keyword" — search across all notesNotes feed section drafts, and drafts must trace every claim to its source in under 30 seconds. Capture in the note:
"exact words" (p. 12)