slidekit-templ
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill interprets text and layout from user-provided PDF files, creating a surface for indirect prompt injection.
- Ingestion points: User-supplied PDF files are converted to slide images which the agent is instructed to analyze.
- Boundary markers: The agent lacks specific delimiters or warnings to ignore instructions that might be visually present in the slide images.
- Capability inventory: The skill can write HTML files to the local disk and execute the pdftoppm command via a Python script.
- Sanitization: Instructions demand high-fidelity reproduction of text content, which may lead to the agent following malicious instructions embedded in the slide content.
- [COMMAND_EXECUTION]: The skill uses subprocess.run to call the pdftoppm utility for image conversion.
- Evidence: The scripts/pdf_to_images.py script executes the system command using an argument list.
- Context: The use of an argument list instead of a raw shell string follows security best practices to prevent command injection. This is a core functionality required for the skill's operation.
- [EXTERNAL_DOWNLOADS]: The skill references CSS and font assets from well-known content delivery networks in its generated HTML output.
- Evidence: Boilerplate HTML includes links to cdn.jsdelivr.net (for Tailwind CSS and Font Awesome) and fonts.googleapis.com.
- Context: These references target reputable services commonly used for hosting web assets and do not involve executable code downloads.
Audit Metadata