菜单
This skill should be used when the user asks to "read this PDF", "extract text from PDF", "what's in this PDF", or provides a PDF file path and wants to understand its contents. Extracts structured text from PDF files using PyMuPDF with pypdf fallback.
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
Use when the user wants to log meeting or call notes for a project — task assignments per student, follow-ups, deadlines. Creates Operon file-tasks in the vault (one .md per task, assigned to the person) and interleaves them on the project hub card via Task Wikilink Overlay. Trigger on: "запиши звонок", "записать встречу", "call notes", "что сделать студентам", "задачи после звонка".
Use when the user wants to set up the Operon task manager in their Obsidian vault the way this lab does it, or reproduce that setup on a fresh machine — flat project tags (no parentTask hierarchy), a "my tasks" table dashboard, service-link badges on project pages, emoji task icons, and day-boundary auto-archiving. Applies plugin settings (data.json), templates, a CSS snippet, optional main.js display patches, and an optional macOS launchd archiver. Trigger on "set up Operon", "настрой Operon как у тебя", "operon setup", "воспроизведи таск-систему Obsidian", "как у тебя задачи и страницы проектов в Obsidian".
This skill should be used when the user wants to write a peer-review of a research paper for any conference or journal, to the standard of a strong reviewer at a top venue. Trigger on "напиши ревью на статью", "write a review", "review this paper", "прогони ревьюера/теоретика/литератора", or when handed a paper PDF to review. Runs a thorough reviewer + theoretician (appendix/proofs) + literature-scout (related work, uncited papers) + experiments-auditor (numbers, hyperparameter search, released code) pass, extracts the user's own PDF annotations, surfaces a weakness skeleton in chat first, then writes one review markdown file per paper in the Summary / Strengths-and-Weaknesses / Questions / Limitations format. Handles prompt injections inside PDFs safely. Venue-agnostic.
Use when the user gives a code repository (a GitHub URL or a local/server path) and wants it added to the Obsidian Code library — a one-time deep analysis that produces a structured set of Markdown notes mapping how the repo works (entrypoint, modules, where to look to change X), without copying the code. Trigger on "add this repo to the code library", "разбери этот репозиторий", "code-ingest <url>", "задокументируй код", or when a project starts using an external repo that is not yet in Code/.
Reference for how the user runs and documents code. Use when an agent needs to understand the user's code workflow — which coding skills and rules apply, how the Obsidian Code library works, how a project hub card records the repo it uses and the project-specific edits, and how experiments are monitored (graphs + tables + loop). Trigger on "how do we work with code here", "что у меня есть для кодинга", "code workflow", before running experiments for a project, or when onboarding to a project that has code.
Use when the user wants to create a new research project, set up a new paper project, initialize a project folder, or says "create project", "new project", "setup project". Creates the standard ~/Papers/<project>/ folder structure with .claude/CLAUDE.md configured for SSH servers and code paths.
| name | pdf-reader |
| description | This skill should be used when the user asks to "read this PDF", "extract text from PDF", "what's in this PDF", or provides a PDF file path and wants to understand its contents. Extracts structured text from PDF files using PyMuPDF with pypdf fallback. |
Extract text from PDF files using open-source tools. PyMuPDF (fitz) is the primary engine with pypdf as fallback.
.pdf file path and asks to read or extract its contentsdefuddle skill instead)Run the extraction script:
python3 ~/.claude/skills/pdf-reader/scripts/extract_pdf.py <pdf_path>
Options:
-p, --pages — extract specific pages, e.g. 1-5 or 1,3,7-m, --method — force fitz or pypdf (default: auto)--json — output structured JSON with metadata--metadata-only — print only title, author, page countExtract in chunks to avoid token overflow:
--metadata-only to check page count-p 1-10, -p 11-20, etc.If extracted text is empty or garbled, the PDF is likely scanned. Inform the user that OCR is needed. Consider suggesting marker or ocrmypdf as next steps.