clip
Pass
Audited by Gen Agent Trust Hub on Sep 17, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The installation instructions include fetching the CLIP library directly from OpenAI's official GitHub repository.
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data in the form of images and text labels during classification and retrieval tasks.
- Ingestion points: Files are ingested via
PIL.Image.openand text is processed viaclip.tokenizeinSKILL.mdandreferences/applications.md. - Boundary markers: None present; the examples assume direct processing of provided data.
- Capability inventory: Code snippets demonstrate image/text encoding and similarity computation using the
torchandcliplibraries. No dangerous system calls or network exfiltration patterns are present. - Sanitization: Standard preprocessing is performed on images (resizing, normalization) as part of the model's pipeline.
Audit Metadata