clipsketch-ai-video-storyboard
Pass
Audited by Gen Agent Trust Hub on Oct 8, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill requires downloading and running code from unverified external sources including: cloning the project repository from https://github.com/RanFeng/clipsketch-ai.git and instructions to run a Docker container using the image earisty/clipsketch-ai:latest from Docker Hub. Neither source is within a verified or trusted organization scope.- [INDIRECT_PROMPT_INJECTION]: The application processes data from external videos and user inputs which are directly interpolated into AI prompts for narrative analysis and storyboard generation.
- Ingestion points: Video frame metadata and string variables such as videoContext and selectedCopy in SKILL.md functions.
- Boundary markers: The prompts lack delimiters or safety instructions to separate untrusted data from the system instructions.
- Capability inventory: The skill performs network operations via the Gemini API and can export generated data to the local file system in ZIP or TXT format.
- Sanitization: There is no evidence of input validation or filtering logic before data is sent to the generative model.
Audit Metadata