clip
Pass
Audited by Gen Agent Trust Hub on Nov 1, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: Fetches the CLIP model implementation from OpenAI's official GitHub repository. It also installs common dependencies like torch, torchvision, and tqdm from standard package registries.
- [INDIRECT_PROMPT_INJECTION]: The skill processes external image files and text labels (SKILL.md, references/applications.md) for classification and search tasks. No explicit boundary markers or sanitization are implemented for this data ingestion. However, the implementation does not include capabilities for dynamic code execution or unauthorized file system modifications using the processed data.
Audit Metadata