The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The documentation provides instructions for installing several well-known third-party libraries via package managers to support PDF operations. These include Python packages like pypdf, pdfplumber, reportlab, pdf2image, and pypdfium2, as well as JavaScript libraries like pdf-lib and pdfjs-dist.
[COMMAND_EXECUTION]: The skill guides the agent to use various command-line utilities for PDF manipulation, including qpdf, pdftk, and the poppler-utils suite (pdftotext, pdftoppm, pdfimages). These are standard tools for the stated purpose.
[PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface as it is designed to ingest and process untrusted PDF documents. Malicious instructions could be embedded within the text or metadata of processed PDFs.
Ingestion points: PDF files are read and processed by multiple scripts (e.g., extract_form_field_info.py, check_fillable_fields.py) and command-line tools.
Boundary markers: No explicit delimiters or warnings are used to isolate extracted text content from instructions.
Capability inventory: The skill includes scripts for writing files, adding annotations, and executing shell-based PDF utilities.
Sanitization: The provided scripts focus on data extraction and transformation; they do not implement specific sanitization or filtering of the extracted text content.
[SAFE]: The script scripts/fill_fillable_fields.py implements a dynamic modification (monkeypatch) of the pypdf library at runtime. This is documented as a legitimate workaround for a known bug in the library's handling of selection lists and does not present a security risk.

pdf