frontend-slides
Pass
Audited by Gen Agent Trust Hub on Oct 29, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted PowerPoint (.pptx) files and user-provided text to generate HTML presentations, creating a vulnerability surface if the input contains malicious instructions. The skill author has included specific mitigation instructions for the agent.
- Ingestion points: PowerPoint files processed via scripts/extract-pptx.py and user text in Phase 1.
- Boundary markers: Instructions in SKILL.md advise treating imported text as untrusted.
- Capability inventory: Subprocess calls for Python script execution, file system writes for HTML and assets.
- Sanitization: Mandatory instructions to escape extracted text and prevent HTML comment breakouts.
- [COMMAND_EXECUTION]: The skill utilizes a provided Python script (scripts/extract-pptx.py) to parse document contents. The script implements security best practices, including safe path resolution to prevent traversal attacks.
- [EXTERNAL_DOWNLOADS]: The skill requires the installation of standard, well-known Python libraries (python-pptx and Pillow) for document and image processing.
- [DYNAMIC_EXECUTION]: The skill generates HTML files containing JavaScript for presentation control and optional inline editing features based on local templates.
Audit Metadata