The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes untrusted PDF files and instructs the agent to analyze their content to determine form field purposes, creating a potential surface for indirect prompt injection where malicious content within a PDF could attempt to influence the agent's logic.
Ingestion points: PDF files provided by users (e.g., scripts/extract_form_field_info.py, scripts/extract_form_structure.py).
Boundary markers: Not present for the internal text content of PDFs.
Capability inventory: The skill has capabilities to write files (scripts/fill_fillable_fields.py, scripts/create_validation_image.py) and execute shell commands (qpdf, pdftotext).
Sanitization: Includes scripts/check_bounding_boxes.py for validating coordinate logic and internal value checks for form fields in scripts/fill_fillable_fields.py.
[COMMAND_EXECUTION]: The documentation and workflow suggest the use of several common system utilities for PDF manipulation, including qpdf, pdftotext, pdftk, and ImageMagick (magick/convert). These tools are used for their intended purposes within the PDF processing domain.
[EXTERNAL_DOWNLOADS]: The skill references and recommends the installation of standard PDF processing libraries such as pypdf, pdfplumber, reportlab, pytesseract, pdf2image, pypdfium2, and JavaScript libraries like pdf-lib and pdfjs-dist.
[DYNAMIC_EXECUTION]: The script scripts/fill_fillable_fields.py implements a runtime monkeypatch of the pypdf library's DictionaryObject.get_inherited method to correct for specific PDF attribute inheritance behaviors during form filling.

pdf