pinecone
Pass
Audited by Gen Agent Trust Hub on Sep 9, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill documents the installation of the
pinecone-clientpackage and references other standard integration libraries such aslangchain-pineconeandllama-index-vector-stores-pinecone. These are official or well-recognized packages for working with the Pinecone service. - [INDIRECT_PROMPT_INJECTION]: The skill facilitates the ingestion of external vector data and metadata into the agent's workflow. This presents a potential surface for indirect prompt injection if the processed data is derived from untrusted sources. 1. Ingestion points: Data ingestion occurs through
upsertandqueryoperations described inSKILL.mdandreferences/deployment.md. 2. Boundary markers: The provided code examples do not include specific delimiters or instructions to ignore embedded commands in the metadata. 3. Capability inventory: The skill utilizes network capabilities to interact with the Pinecone API and provides tools for data storage and retrieval. 4. Sanitization: The documentation does not explicitly detail sanitization procedures for the metadata or vector values before they are processed by the agent.
Audit Metadata