video-description-oversight
Pass
Audited by Gen Agent Trust Hub on Aug 20, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
- [SAFE]: The skill consists entirely of documentation and process guidelines. It does not contain any executable code, scripts, or tool definitions that could be used for malicious purposes.
- [EXTERNAL_DOWNLOADS]: The skill references several research papers and public code repositories on platforms such as ArXiv, GitHub, and Hugging Face. These are informational references only and do not involve runtime downloads or execution of untrusted code.
- [PROMPT_INJECTION]: The skill outlines a process for reviewing external video captions, which constitutes an ingestion surface for indirect prompt injection. However, the workflow mitigates this risk through a structured critique methodology involving specific aspects (Subject, Scene, Motion, Spatial, Camera) and explicit acceptance criteria, preventing the data from being interpreted as agent instructions.
Audit Metadata