PDF Utils
Use this skill for local, scriptable PDF processing. It is a stable 1.x skill for OCR, arXiv reference mining, and repeatable PyMuPDF workflows. Prefer the built-in pdf tool for AI-style reading, summarization, question-answering, and semantic analysis of PDF content.
Choose the right tool
- Use the built-in
pdftool for summary, Q&A, extraction by meaning, or general document understanding. - Use
scripts/extract_refs.pywhen the PDF already has extractable text and you need arXiv IDs or batch downloads. - Use
scripts/ocr_pdf.pywhen the PDF is scanned/image-based and text extraction is poor or empty. - Use
scripts/pdf_ops.pyfor repeatable local PDF operations such as merge, split, and rendering a page to an image.
Core workflows
Extract arXiv IDs from a text PDF
Run:
python3 scripts/extract_refs.py paper.pdf
If needed, download the referenced papers:
python3 scripts/extract_refs.py paper.pdf --download --out ~/papers/
OCR a scanned PDF
Run OCR on all pages:
python3 scripts/ocr_pdf.py paper.pdf --all
To OCR and immediately extract arXiv IDs from the OCR output:
python3 scripts/ocr_pdf.py paper.pdf --all --extract-refs
Dependencies
Install these before using OCR features:
brew install tesseract
brew install tesseract-lang
pip3 install pytesseract Pillow pymupdf --break-system-packages
Read more only if needed
- Read
references/usage.mdfor CLI examples, programmatic API notes, PDF ops usage, and known limits. - Read the scripts directly if you need to patch behavior or reuse helper functions.
Practical guidance
- For very large PDFs, OCR in page ranges or batches instead of all at once.
- For handwritten or low-resolution scans, expect OCR quality to drop.
- If a PDF yields partial references, inspect the reference pages first instead of assuming extraction is complete.
- For merge/split/page rendering, use
scripts/pdf_ops.pyfirst before writing one-off snippets.
Related skills
- Downloads
- 1.1kAll time
- Security audit
- Review
- Last updated
- 6mo ago
- Current version
- v1.0.1
- License
- MIT-0
