MinerU Skill: PDF to Markdown

Nebutra/MinerU-Skill

Parse PDFs, scans and Office files into Markdown with MinerU, with OCR, formula and table extraction, batch processing and optional local parsing.

内容更新日:

GitHub で見る
npx skills add Nebutra/MinerU-Skill
133スター6フォークリポジトリの最終プッシュ

このスキルについて

Use a Python CLI to parse local files or public document URLs and return Markdown to an agent. The default cloud route uses MinerU services; an optional local engine handles supported PDFs with an existing text layer. Root and Claude-plugin entrypoints describe the same mineru workflow, rather than separate catalog tasks.

おすすめする理由

An agent-oriented CLI around MinerU, adding Markdown stdout, machine-readable status, resumable batches and explicit cloud/local modes.

役立つこと

  • Extract readable text, tables and formulas for a document workflow.
  • Process a folder of documents with resumable batches.
  • Return Markdown directly to an agent or write a local knowledge folder.

Workflow

Choose the processing route

Decide whether the input needs cloud OCR or the optional local text-PDF engine.

Parse and inspect status

Run the CLI with --stdout for Markdown or --json for machine-readable status; use --resume for repeated batches.

Check extracted content

Inspect tables, formulas and images against source pages before reusing the Markdown.

Author-maintained Skill instructions

Installation and runtime

Install the Skill instructions, then provision the documented runtime dependencies before running a task.

コマンド用途と使いどころ
python3 scripts/mineru.py paper.pdf --stdoutReturn parsed Markdown to the terminal.
python3 scripts/mineru.py paper.pdf --jsonReturn machine-readable parsing status.
python3 scripts/mineru.py --doctorCheck the local environment and optional integrations.

Author-maintained Skill instructionsOfficial project README

Scope and practical limits

This is a third-party workflow around MinerU, separate from the already-listed MinerU engine. Cloud OCR and local text parsing are different execution paths.

Author-maintained Skill instructionsOfficial project README

License and source

The project is distributed under MIT. Agent subscriptions, external services and optional integrations have their own requirements.

MIT license

使う前に

  • Python 3.8+ for the standard-library CLI; optional local parsing, splitting and integrations need their own extras.
  • Default cloud processing transfers documents to MinerU. Tokens, quotas and supported formats depend on the provider and selected route.
  • The token-free Agent route and token-based Standard route have different limits. Verify the current provider conditions before large jobs.

対応エージェント

Claude CodeCodex CLICursorHermes AgentGoogle Antigravity

資料・参照元

元の資料は以下のリンクから確認できます。実践のヒントは SkillHey による編集上の提案です。