working-with-documents

Fail

Audited by Gen Agent Trust Hub on Feb 16, 2026

Risk Level: HIGHPROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [PROMPT_INJECTION] (HIGH): The skill is designed to ingest data from untrusted external sources, creating a high risk for indirect prompt injection.
  • Ingestion points: Content extraction is performed via pandoc, pdfplumber, pdftotext, markitdown, and pytesseract (SKILL.md).
  • Boundary markers: No delimiters or isolation instructions are present to differentiate document content from system instructions.
  • Capability inventory: The skill can execute complex system commands (soffice, qpdf, pandoc), modify the filesystem, and perform OCR, providing significant impact for any successful injection.
  • Sanitization: There is no evidence of filtering or sanitization of the extracted text before it is processed by the agent.
  • [COMMAND_EXECUTION] (MEDIUM): The skill relies extensively on shell command execution for utilities like LibreOffice (soffice), qpdf, and pandoc. This presents a risk of command injection if file paths or extracted content are improperly handled during command construction.
  • [EXTERNAL_DOWNLOADS] (LOW): The skill requires a large number of Python and Node.js packages, as well as system-level utilities (apt-get install). While these are from established repositories, they expand the overall attack surface and require elevated privileges for setup.
Recommendations
  • AI detected serious security threats
Audit Metadata
Risk Level
HIGH
Analyzed
Feb 16, 2026, 03:40 AM