Skip to main content
skill CX Codex

extract-figures — Claude Code skill: DOCX figures

extract-figures pulls figures, charts, diagrams, and images from .docx, converts EMF/WMF assets to PNG, and prepares LaTeX usage notes.

no license ★ 240 updated 2026-09-18

Install into Claude Code · Skills CLI (Vercel) All methods and platforms ↓
npx skills add franklee16/academic-research-skills --skill extract-figures -a claude-code
Open on GitHub

What extract-figures does

This skill exports embedded media from a Word document into a figures/ folder beside the source .docx. EMF, WMF, EMZ, and WMZ files become 300 DPI PNGs; PNG, JPG, GIF, and BMP assets are copied in their existing formats.

The bundled scripts/extract_figures.py script opens the DOCX archive, maps relationship IDs to media files, walks the document XML, and looks for nearby captions such as “Figure N”. It writes position-based names like img_001.png, a _manifest.json mapping, and a draft _readme_draft.md file.

The workflow includes checking image content when a figure has multiple panels, then correcting filenames and the manifest if visual order differs from XML order. Python plus python-docx and Pillow are required before the skill can run.

What it helps with

  • export every image from a .docx file
  • convert EMF and WMF artwork for LaTeX
  • match extracted files with Figure N captions
  • check panel order in compound figures
  • prepare a README with LaTeX snippets

Example prompts for the agent

  • Extract every figure, chart, and image from paper.docx into figures/.
  • Pull the diagrams from manuscript.docx, convert vector assets to PNG, and create LaTeX usage notes.
  • Check the panel order in figures from Draft/results.docx and update the manifest filenames.

How to install extract-figures

1Skills CLI (Vercel)

npx skills add franklee16/academic-research-skills --skill extract-figures -a claude-code

2Manually: put the files in place

~/.claude/skills/extract-figures/SKILL.md

Copy the whole skill folder; for one project use .claude/skills/extract-figures/

Step by step, with commands for each platform: How to install skills in Claude Code: five ways

Questions about extract-figures

How do I install extract-figures in Claude Code?

Run npx skills add franklee16/academic-research-skills --skill extract-figures -a claude-code, then check ~/.claude/skills/extract-figures/SKILL.md.

What does extract-figures do?

It exports embedded media from .docx files, turns EMF and WMF assets into 300 DPI PNGs, associates images with nearby captions, and prepares LaTeX notes.

What does extract-figures need to run?

It requires Python with python-docx and Pillow installed. Install them with pip install python-docx Pillow.

Similar skills

All skills for Office documents →

From the same set98

The whole set →
consistency-checker consistency-checker runs a pre-submission consistency check on finance and accounting manuscripts before journal submission or revision. CX Codex +1 ★ 240repo stars quality-checks-diagnosis quality-checks-diagnosis helps choose the right survey check and interpret failures. Use it when you need to distinguish inattention from unclear instructions or a weak experimental treatment. CX Codex +1 ★ 240repo stars avoid-leading-questions avoid-leading-questions reviews a draft survey question, names the bias behind it, and gives a concrete rewrite for each of the three cases it covers. CX Codex +1 ★ 240repo stars dril-dataset-construction dril-dataset-construction implements the DRIL methodology to assemble economic datasets from primary sources via AI agents. Ideal for constructing panel data, legal variables, and cited codebooks. CX Codex +1 ★ 240repo stars aer-referee-sim aer-referee-sim runs a finished manuscript through a mock AER desk screen and three referees, then returns a verdict and a prioritized revise list. CX Codex +1 ★ 240repo stars practitioner-finance-voice The practitioner-finance-voice skill rewrites academic finance drafts into the direct tone of practitioner outlets like FAJ and JPM. It frames findings for portfolio managers without fabricating facts. CX Codex +1 ★ 240repo stars comments-synthesis comments-synthesis turns feedback from several discussants, referees or workshop readers of one paper into a single revision plan, with the points they repeat ranked first. CX Codex +1 ★ 240repo stars nature-figure nature-figure structures high-impact publication figure design using Python or R backends. CX Codex +1 ★ 240repo stars

Project sponsors

Companies that keep this analytics open