office-pptx
Pass
Audited by Gen Agent Trust Hub on Aug 25, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill utilizes local shell commands to perform necessary media processing and file management tasks. Specifically, it executes the
icaclsutility on Windows systems to ensure generated PPTX artifacts have the correct read permissions for the user. It also usesffprobeto extract duration metadata from narration audio files. These operations are strictly limited to the skill's primary purpose of creating and preparing presentation artifacts. - Evidence found in
scripts/svg_to_pptx/pptx_package/builder.pyandscripts/svg_to_pptx/pptx_package/narration.py. - [PROMPT_INJECTION]: The skill provides instructions for the agent to extract text from existing PowerPoint presentations, which introduces a surface for indirect prompt injection. If an untrusted presentation contains malicious instructions, they could be ingested into the agent's context during the extraction process.
- Ingestion points: Text extraction from presentation shapes via the provided
python-pptxcode snippet inSKILL.md. - Boundary markers: None are implemented in the provided code snippet.
- Capability inventory: The skill includes scripts for file creation and modification, as well as local command execution (
icacls,ffprobe). - Sanitization: Extracted text is processed and printed to context without filtering or sanitization.
Audit Metadata