The Agent Skills Directory

[PROMPT_INJECTION]: The skill includes functionality for reading and extracting text from PDF documents using libraries like pypdf and pdfplumber. While standard, this introduces an indirect prompt injection surface where malicious instructions embedded in a processed PDF could potentially influence the agent's behavior. No active injection or safety bypass instructions were found.\n
Ingestion points: scripts/extract_form_field_info.py, scripts/extract_form_structure.py, and reference.md.\n
Boundary markers: None explicitly mentioned in the processing instructions.\n
Capability inventory: The skill uses pypdf for file writing and provides instructions for executing command-line processing tools.\n
Sanitization: None implemented, as the scripts are designed for direct utility processing.\n- [DYNAMIC_EXECUTION]: The script scripts/fill_fillable_fields.py utilizes runtime monkeypatching to modify the pypdf library's internal behavior for handling choice fields. This modification is limited to ensuring compatibility and does not present an execution risk.\n- [COMMAND_EXECUTION]: The documentation provides examples and instructions for executing common command-line utilities such as qpdf, pdftk, and magick. These are standard tools for the operations described and represent legitimate use within the skill's scope.

pdf-processing-anthropic