pyvespa
Pass
Audited by Gen Agent Trust Hub on Sep 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill facilitates the processing of user-supplied data through document feeding and query operations, representing a standard attack surface.
- Ingestion points: Untrusted data enters the agent context via document field values in feeding operations and query parameters in the YQL DSL across all referenced documentation.
- Boundary markers: The instructions do not demonstrate the use of explicit delimiters or boundary markers to isolate user data from command structures.
- Capability inventory: The skill performs network requests to Vespa endpoints and manages local infrastructure through the Docker SDK.
- Sanitization: Input validation is assumed to be handled by the client library and backend service rather than specified in the skill instructions.
- [EXTERNAL_DOWNLOADS]: The skill refers to the installation of official libraries and the retrieval of assets from trusted sources.
- References the installation of the
pyvespapackage from the official Python Package Index. - Includes configuration examples in
docs/package-api.mdfor downloading models and tokenizers from Hugging Face's official repository for embedder integration.
Audit Metadata