The Agent Skills Directory

[Indirect Prompt Injection] (LOW): The skill ingests untrusted user stories and scripts to generate prompts for media creation tools.
Ingestion points: User-provided story ideas, scripts, and conceptual descriptions (Step 1).
Boundary markers: Absent. The instructions do not specify delimiters or warnings to ignore embedded instructions in the user's input.
Capability inventory: Access to draw_one_image and draw_one_video tools (Steps 3 and 4).
Sanitization: The skill provides guidance on escaping JavaScript strings to prevent UI breakage, but does not provide sanitization or filtering for the natural language prompts sent to the image/video generation engines.
[Data Exposure & Exfiltration] (LOW): The documentation suggests using picsum.photos for placeholder images in the generated storyboard HTML.
Evidence: The "Placeholder Image Service" section recommends https://picsum.photos/seed/... as an external source. While benign for this use-case, it is a non-whitelisted domain used for network requests in the output.
[Dynamic Execution] (LOW): The troubleshooting section (FAQ) contains a Node.js snippet that uses new Function() to validate JavaScript syntax in the generated HTML.
Evidence: The shell command node -e "... new Function(scriptMatch[1]); ..." performs a check on the generated storyboard.html. If an agent were to execute this automatically on a file containing unescaped user-controlled descriptions, it could lead to code execution in the troubleshooting context.

storyboard-generator