pdf-converter skill (tanis90) for Claude Code: turn PDFs into Markdown, Word or LaTeX

Third-party skill that has Claude convert PDFs, scans and Office files to Markdown, Word, HTML or LaTeX with the MinerU command line tool.

Installs: 9.6K; on skills.sh
Stars: 55; on GitHub
Updated: May 2026; last commit

pdf-converter makes Claude run the MinerU command line tool (mineru-open-api) on a PDF, image or Office file and turn it into Markdown. Claude then reads the result to answer your question, pull out tables or summarize.

It has two modes. flash-extract needs no sign-in and handles short files. extract needs a MinerU token and handles bigger files, more output formats and batches. Both send your document to MinerU's service at mineru.net, so check that you are happy with that before you use it on private files.

Good for

Anyone who wants Claude to read, summarize or quote a long PDF
People who need tables or formulas pulled out of a paper
Teams converting batches of PDFs or Office files to Word, HTML or LaTeX
Readers who need OCR for scanned documents

Install pdf-converter

Claude Code

Copy this prompt.

Install the pdf-converter skill in Claude Code. Run this command:

npx skills add tanis90/pdf-converter-mineru

Then check that it loaded and tell me how to use it.

2
Paste it into Claude Code. It shows the change and asks before running it.

Requires Node.js.

The skill calls the mineru-open-api command, so install it too: pick one of the four MinerU CLI methods SKILL.md lists. flash-extract needs no sign-in. For extract, run mineru-open-api auth once. The two script routes pipe a downloaded script into your shell; the npm or uv route is the milder choice.

From the project README. Source.

What's in the skill

When it loads

PDF converter powered by MinerU — convert PDF to Word, Markdown, HTML, LaTeX, or plain text. Also handles image-to-text OCR, scanned document recognition, and Office formats (DOCX, PPTX, Excel). Supports 80+ languages. Use this skill when the user wants to convert, extract, read, parse, or summarize any PDF or document. Also applies when the user shares a PDF file or link and asks about its content, needs tables or formulas extracted, wants PDF OCR, or says things like 'turn this into a doc' or 'what does this paper say'.

Sections in SKILL.md

Language Rule
Core Workflow
Page Range Extraction Rule
Two Extraction Modes
flash-extract — Fast, no auth
extract — Precision, auth required
When to Use Which
Language Support
Data Flow
Troubleshooting

Key facts

Author: tanis90 (the repo owner); the skill uses MinerU from OpenDataLab (Shanghai AI Lab)
License: None stated: the repo has no LICENSE file
Installs on skills.sh: 9,638 as of 2026-11-02
GitHub stars: 55 as of 2026-11-02
What is inside: 3 files in the repo: SKILL.md, README.md and .gitignore. It bundles no scripts.
Last commit: 2026-05-15
Needs: The mineru-open-api command line tool, installed with npm, uv or an install script
Works in: Claude Code (the README calls it a Claude Code skill)

What it includes

Quick conversion with no sign-in, up to 10 MB or 20 pages
Runs mineru-open-api flash-extract on a PDF, image, DOCX, PPTX or Excel file, or on a URL. The output is Markdown only, and images, tables and formulas may become placeholders.
Try: “What does report.pdf say? Give me a summary.”
Runs code · flash-extract
Full-fidelity conversion, up to 200 MB or 600 pages, needs a token
Runs mineru-open-api extract for preserved images, accurate tables, formulas and batches. Output can be md, json, html, latex or docx, chosen with -f. The --ocr flag forces OCR for scanned files.
Try: “Convert report.pdf to Word and keep the tables.”
Runs code · extract
Sets up the MinerU token that extract needs
If you have not signed in yet, Claude points you to this command before it uses extract.
Runs code · mineru-open-api auth
Claude reads the Markdown and does the work you asked for
MinerU only converts. Claude then summarizes, extracts tables or lists references from the output. For quick questions it reads stdout. For conversions it saves a file with -o.
Read · Read and process step
Each page chunk gets its own output file name
When you split a file by pages, Claude must write each chunk to an explicit file name such as report_p1-20.md. Otherwise chunks overwrite each other in the output folder.
Write · Page range rule
--language picks the OCR language; the default is Chinese plus English
The skill lists codes such as en, fr, de, japan, korean and es, and says 80+ languages work in total. It also tells Claude to reply in the language you write in.
Runs code · Language codes
-v shows HTTP details and --timeout 600 helps with big files
The skill also lists four ways to install the CLI if it is missing.
Runs code · Troubleshooting flags

From the official docs

Permissions

Medium risk

Your documents are sent to an outside service for conversion, and the skill can install a command line tool on your machine. Use it on files you are happy to upload.

What it can do

Read your data
Create and change things
Delete things
Read-only mode

Watch out for

Your documents are uploaded to mineru.net
Both modes send the file to MinerU's service. Do not use it on confidential PDFs unless you accept that. github.com
The CLI install can run a remote script
SKILL.md lists curl -fsSL ... | sh on macOS and Linux and irm ... | iex on Windows, besides npm i -g and uv tool install. Prefer the npm or uv route, and read a script before you pipe it into a shell. github.com
It can write files to your disk
With -o, Claude saves converted files and extracted assets in the folder you name. Name an empty output folder so nothing gets overwritten. github.com

Security advisories

No published security advisories as of Nov 2, 2026.

Technical details

Talks to	mineru.net (documents are sent here for processing). cdn-mineru.openxlab.org.cn (hosts the install scripts SKILL.md lists for the CLI)
Sign-in and keys	flash-extract needs no key. extract needs a token you create with mineru-open-api auth; the skill does not say where the CLI keeps it.
Data retention	SKILL.md says both commands send the document to MinerU's API at mineru.net and calls it "a stateless API call with no persistent storage". That is the skill author's description, not MinerU's own policy.
Scopes	Runs mineru-open-api on files you name, and on URLs you give it Writes converted files to the folder you choose with -o Can install the mineru-open-api CLI when it is missing

Questions

What does the pdf-converter skill do?

It has Claude convert PDFs, images and Office files to Markdown with the MinerU command line tool. Then Claude reads the result to summarize it, pull out tables or find references. extract mode can also write Word, HTML, LaTeX or JSON.

How do I install pdf-converter?

Use the install box at the top of this page, then install the mineru-open-api tool the skill calls. The box lists the four install options from SKILL.md. The README only documents the skills CLI step.

Does it need an API key?

Not for flash-extract, which SKILL.md describes as "Fast, no auth" with a limit of 10 MB or 20 pages. extract needs a token created with mineru-open-api auth, and it allows 200 MB or 600 pages.

Is pdf-converter safe to use on private files?

Think first. Both modes send the document to MinerU's API at mineru.net, and SKILL.md says the call keeps no files. That is the skill author's statement, so read MinerU's own terms for sensitive material. Our advice: use it on public or low-risk documents.

Does it work in the Claude app or Codex?

The README calls it a Claude Code skill and does not name the Claude app or Codex. Our advice: use Claude Code, because the skill runs a local command line tool, so no Claude app upload option is offered here. The install box lists the generic Codex skills folder.

Is this the official MinerU skill?

No. The repo belongs to tanis90, and the README says the skill is powered by MinerU Open API from OpenDataLab.

Information

updated Nov 2, 2026

Repository: tanis90/pdf-converter-mineru

tanis90/pdf-converter-mineru skills.sh

Sources

Docs, issue trackers and discussions this page is built from.

Show all 2

README.md · github.com
SKILL.md · github.com