| name | pdf-reader |
| description | Read, search, extract, or render PDFs, including page-specific or visual questions. |
PDF Reader
Routing card for hybrid PDF text extraction and rendered-page reading.
Commands
Use uv; do not create a persistent virtualenv.
uv run --with pymupdf SKILL_DIR/scripts/pdf_info.py <path>
uv run --with pymupdf SKILL_DIR/scripts/pdf_extract.py <path> --pages 1-5
uv run --with pymupdf SKILL_DIR/scripts/pdf_search.py <path> "theorem 3.2" --context 3
uv run --with pymupdf SKILL_DIR/scripts/pdf_render.py <path> --pages 3 --dpi 150
Page specs are 1-indexed: all, 1-5, 1,3,7, or 3.
Strategy
- Triage first with
pdf_info.py.
- Search/extract before rendering when text is enough.
- Render pages when equations, diagrams, layout, or scanned content matter; read PNGs with the
read tool.
- For long PDFs, target sections/pages instead of full-document rendering unless explicitly requested.
- Cite page numbers in answers.