Figure Extractor
Extract complete, publication-quality figures and images from documents — local PDFs, PDF URLs, arXiv links, and HTML article pages. Use whenever the user wants to pull figures / charts / diagrams / images out of a paper or web article, says "extract the figures from this PDF / arXiv paper / webpage", needs clean figure PNGs for slides, decks, or reuse, or finds that naive embedded-image extraction returns broken fragments. This is a command-line tool (Python + PyMuPDF) that must be run in a shell; if your runtime has no shell, see the degraded-mode fallback. Keywords: extract figures from pdf
Install
npx skills add Sunrich-HT/figure-extractor --skill figure-extractorgit clone --depth 1 https://github.com/Sunrich-HT/figure-extractor.git /tmp/figure-extractor
mkdir -p ~/.claude/skills
cp -r /tmp/figure-extractor/. ~/.claude/skills/figure-extractorPersonal skills live in ~/.claude/skills. Use .claude/skills in a project folder to share a skill with your team.
From the SKILL.md
# Figure Extractor Skill Use this skill when the user needs complete figures or images extracted from PDFs, arXiv papers, or HTML article pages. ## What this skill actually is (read first) `figure-extractor` is a **command-line tool** (Python + PyMuPDF). "Using it" means running it in a shell — e.g. `figure-extractor extract paper.pdf`. It is **not** an in-agent callable, a hosted service, or a "page" you activate. It does **not** require installation or network. **PDF and HTML sources both** need a Python interpreter with **PyMuPDF and nothing else** — no pip, no PyPI, no GitHub, no Pillow, no beautifulsoup4. If you cannot install packages, use the single-file build, or the reduced extractor printed in this file (Step 0 below). Only URL sources need network. If the user gives you a **browser PDF-viewer URL** (`.../viewer.html?file=...`, Chrome's built-in viewer, Google Docs viewer), pass it in as-is: the real document is unwrapped from `?file=` / `#file=` / `?url=` before anything is fetched. If they give you a **saved HTML page**, pass the local file: images stored as inline `data:` URIs are decoded without touching the network. **Invocation contract** - **Mechanism:** a shell command — installed CLI, `python -m figure_extractor`, or the single file. - **Input:** a local PDF/HTML path or a PDF/arXiv/HTML URL. - **Output:** PNG files + `manifest.json` + `contact_sheet.jpg` (+ optional `figures.zip`) written to `--out`.
Details
- Skill name
figure-extractor- Repository
- Sunrich-HT/figure-extractor
- License
- MIT
- Repo updated
- 19 August 2026
FAQ
What does the Figure Extractor skill do?
Extract complete, publication-quality figures and images from documents — local PDFs, PDF URLs, arXiv links, and HTML article pages. Use whenever the user wants to pull figures / charts / diagrams / images out of a paper or web article, says "extract the figures from this PDF / arXiv paper / webpage", needs clean figure PNGs for slides, decks, or reuse, or finds that naive embedded-image extraction returns broken fragments. This is a command-line tool (Python + PyMuPDF) that must be run in a shell; if your runtime has no shell, see the degraded-mode fallback. Keywords: extract figures from pdf
How do I install the Figure Extractor skill?
skills CLI (Claude Code, Cursor, Codex and more): npx skills add Sunrich-HT/figure-extractor --skill figure-extractor. Skills go in ~/.claude/skills (personal) or .claude/skills (project), and Claude loads them automatically when a task matches the description.
Is Figure Extractor popular?
It lives in Sunrich-HT/figure-extractor, which has 5 GitHub stars and 0 forks.
Related skills
Generating PPTX From Template
5anikch/ppt-from-template-skill
Generates fully populated PowerPoint presentations that inherit the exact colors, fonts, layouts, and branding from a user-supplied .pptx template. Use when the user uploads or references a .pptx template file and asks to create or generate a presentation on a given topic. Triggers on: use my template, create slides from my pptx, generate a deck using my template, fill in my template, make a presentation from the attached file, build slides using my brand template, or when the user shares a .pptx file and wants a polished output matching their brand. Also use when multiple .pptx files are uplo
DOCX Formatter
5CP-here/docx-formatter-skill
将文本或 Markdown 内容排版为专业的中文 Word (.docx) 文档,支持 OMML 数学公式、表格框图、标准排版、引用上标和轻量完整性验证。当用户要求 Word 排版、把 Markdown 或纯文本转成 Word 文档、生成带公式的 Word 文档、用数学方程格式化文档、在 Word 中创建流程图样式图表、将参考文献引用渲染为上标、或格式化文档为 Word 时调用。
Kindlepub
5jessl2juice/kindlepub
End-to-end Amazon KDP self-publishing pipeline for a finished book manuscript. Builds the reflowable Kindle EPUB (epubcheck-clean), generates cover art with Gemini Nano Banana and assembles the cover in KDP Cover Creator (which auto-sizes the spine — no custom print PDF), drafts all KDP metadata, and walks the user through the KDP Details/Content/Pricing/Publish flow for Kindle ebook, paperback, and hardcover. Use this whenever the user wants to publish, self-publish, or upload a book to Amazon, KDP, or Kindle Direct Publishing; make a Kindle ebook; set up a paperback or hardcover; build a boo
Diorama Journey
5nocoo/diorama-journey
Create narrated story videos through connected miniature 3D worlds, with one recurring character, object or signal carrying the story across chapters. Use for isometric diorama explainers, process journeys, project stories or narrated retrospectives that need 3D animation, chapter clips and matching presentation slides. Includes a Remotion and Three.js starter, voice generation, PPTX/PDF/ODP exports and production archives.
Ppt Rebuild
5Enei7/ppt-rebuild-skills
Rebuild existing PDF slide decks, especially mathematical lectures, as visually faithful and object-editable PowerPoint with native editable Office equations, calibrated math/text alignment, and exact source-figure crops. Not for creating a new image-only presentation from an outline.
Deck Kit
5kiselas/every-slide-is-code
Presentation decks as code — pitch decks, product launches, idea pitches, board and quarterly reviews, tech talks. One self-contained HTML file with animated charts, diagrams, builds and cinematic transitions (camera moves, morph, zoom, iris, tear, dissolve), a presenter view, and static PDF/PPTX exports. Use whenever someone wants to present an idea, product, plan or result as slides, or asks for a deck, keynote, pitch or presentation that should look designed rather than templated.
Browse the most popular Claude skills, all documents & office files skills, or write your own with the skills generator.