The Agent Skills Directory

Prompt Injection (LOW): The skill is susceptible to indirect prompt injection because it extracts and analyzes text from untrusted PDF files.\n
Ingestion points: scripts/extract_form_field_info.py, scripts/check_fillable_fields.py, and extraction examples in SKILL.md.\n
Boundary markers: Absent. The skill does not implement delimiters or 'ignore' instructions for the extracted text before it is processed by the agent.\n
Capability inventory: Extensive PDF manipulation capabilities including file writing (PdfWriter), metadata extraction, and form filling.\n
Sanitization: Absent. Text extracted via OCR (pytesseract) or direct reading (pypdf, pdfplumber) is passed to the agent without filtering.\n- Dynamic Execution (LOW): scripts/fill_fillable_fields.py performs runtime monkeypatching of the pypdf library (DictionaryObject.get_inherited) to resolve a known bug in selection list handling. This is a targeted compatibility fix.\n- Command Execution (LOW): The SKILL.md guide documents the use of standard system binaries such as qpdf, pdftk, and pdftotext. While these are legitimate utilities, executing them on untrusted files increases the environment's attack surface.

pdf