document-pptx

Pass

Audited by Gen Agent Trust Hub on Oct 6, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONDYNAMIC_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes data from external PowerPoint (.pptx) files, creating a surface for untrusted instructions to influence agent behavior.
  • Ingestion points: Documented in scripts/pptx_inventory.py, scripts/pptx_extract_notes.py, and references/extraction-stack.md.
  • Boundary markers: The provided extraction scripts do not implement delimiters or safety warnings to distinguish between content and instructions.
  • Capability inventory: The skill has access to Bash execution and file writing, and provides templates that use Python's subprocess for system calls.
  • Sanitization: No sanitization is performed on the text extracted from slide components.
  • [COMMAND_EXECUTION]: The skill includes code templates that use the subprocess module to call system utilities for file conversion.
  • Evidence: Found in assets/pitch-deck.md (calling unoconv) and references/pptx-speaker-notes-delivery.md (suggesting libreoffice).
  • [DYNAMIC_EXECUTION]: The skill provides Python scripts and templates that are intended to be executed at runtime to generate or modify presentation files.
  • Evidence: Python logic in assets/pitch-deck.md and assets/quarterly-review.md demonstrates runtime slide assembly and data processing.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 6, 2026, 02:03 PM