| name | readpdf |
| description | Extract embedded text and rendered page images from PDFs. Use when the user asks to read, inspect, parse, convert, or extract contents from PDF files without OCR. |
Read PDF
Use the bundled readpdf.sh script to extract embedded text from PDFs with pdftotext and render each page as a JPEG with pdftoppm.
This skill does not OCR scanned PDFs. It only extracts text already present in the PDF, plus page images for visual inspection.
Commands
Extract to a chosen directory
./readpdf.sh extract -o <output-dir> <pdf-or-dir>...
Inputs may be PDF files or directories. Directories are searched recursively for *.pdf files.
For each PDF, the script writes:
<output-dir>/XXX.txt — embedded text extracted with pdftotext -layout<output-dir>/XXX.N.jpg — rendered page N, at 200 DPI
The script emits every generated file path to stdout.
Read PDFs into model context
./readpdf.sh read <pdf-or-dir>...
This extracts to a temporary directory, emits all generated paths to stdout, and prints a reminder on stderr.
When using read, after running the script you must read the generated .txt files and inspect/read every .jpg page image into context. Prefer reading .txt first, then the images in page order, so both embedded text and visual layout are available.
Notes
- Requires Poppler tools on
PATH: pdftotext and pdftoppm.
- Poppler is installed by Homebrew as
poppler in install.bash.
- If a PDF contains only scanned images, the
.txt file may be empty; still inspect the rendered page images.
- If multiple PDFs have the same basename, later outputs get numeric suffixes to avoid overwriting.
