youtube-render-pdf
Pass
Audited by Gen Agent Trust Hub on Sep 29, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from external sources, specifically YouTube subtitles, metadata, and video frame content.\n
- Ingestion points: Subtitle tracks, video metadata (title, chapters), and visual content from video frames.\n
- Boundary markers: There are no explicit instructions or delimiters used to separate untrusted external content from the agent's instructions, nor are there warnings to ignore instructions embedded in the source video data.\n
- Capability inventory: The agent has the ability to write local files (.tex, .pdf, .py), download images, and execute system commands for image processing and LaTeX compilation.\n
- Sanitization: The instructions do not define a process for sanitizing or escaping the data retrieved from YouTube before it is used in document generation or script execution.\n- [DYNAMIC_EXECUTION]: The skill generates and executes Python scripts at runtime to produce visualizations using
matplotlibandseaborn. While this is a primary feature for creating pedagogical notes, executing dynamically generated code carries inherent risks if the logic is influenced by untrusted input data.\n- [COMMAND_EXECUTION]: The skill relies on executing external shell commands to process video frames (e.g., usingmagick montagefor contact sheets) and to compile the generated LaTeX source into a final PDF deliverable.
Audit Metadata