pdf

Pass

Audited by Gen Agent Trust Hub on Jun 15, 2026

Risk Level: SAFEPROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [PROMPT_INJECTION]: Indirect Prompt Injection Surface. The skill is designed to ingest and process external PDF files, which provides a surface for untrusted data to enter the agent's context. Maliciously crafted PDFs could contain hidden instructions intended to influence the agent's behavior after text extraction.
  • Ingestion points: PDF content is read using pypdf, pdfplumber, and pdf2image across multiple scripts, such as extract_form_structure.py and scripts/extract_form_field_info.py.
  • Boundary markers: No explicit delimiters or instructions are used to signal to the agent that extracted text should be treated as untrusted data.
  • Capability inventory: The skill provides the agent with multiple scripts and command-line tools (e.g., qpdf, pdftotext, magick) capable of reading from and writing to the local file system.
  • Sanitization: The extracted text is not sanitized or validated before being presented to the agent for processing.
  • [EXTERNAL_DOWNLOADS]: Dependency on well-known external libraries. The skill documentation and scripts reference several external dependencies that must be installed for full functionality.
  • Python packages: pypdf, pdfplumber, reportlab, pandas, pytesseract, pdf2image, and pypdfium2.
  • JavaScript libraries: pdf-lib and pdfjs-dist are mentioned in the advanced reference.
  • System utilities: The skill recommends the use of poppler-utils, qpdf, pdftk, and ImageMagick (magick).
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 15, 2026, 09:37 AM
Security Audit — agent-trust-hub — pdf