pptx

Pass

Audited by Gen Agent Trust Hub on Jun 30, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill invokes external system binaries to perform file format conversions and validations.
  • 'soffice' (LibreOffice) is used to convert presentations to PDF and HTML for visual analysis and structural validation.
  • 'pdftoppm' (Poppler) is used to generate slide thumbnails from PDF files.
  • 'git diff' is used to compare document versions during track-changes validation.
  • These commands are called using 'subprocess.run' with list-style arguments, which prevents shell injection.
  • [PROMPT_INJECTION]: The skill creates a surface for indirect prompt injection by extracting text content from user-provided PowerPoint files.
  • Ingestion points: 'scripts/inventory.py' extracts text from shapes; 'ooxml/scripts/unpack.py' extracts raw XML.
  • Boundary markers: Absent. Extracted text content is provided to the agent context without specific delimiters or instructions to ignore embedded instructions.
  • Capability inventory: The skill has the ability to execute shell commands via 'subprocess' (for conversion utilities) and launch a headless browser via 'playwright'.
  • Sanitization: The skill implements mitigation for XML External Entity (XXE) attacks by using the 'defusedxml' library for all XML parsing operations.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 30, 2026, 03:47 PM
Security Audit — agent-trust-hub — pptx