pinecone-query

Pass

Audited by Gen Agent Trust Hub on Feb 28, 2026

Risk Level: SAFEPROMPT_INJECTIONNO_CODE
Full Analysis
  • [PROMPT_INJECTION]: The skill demonstrates a surface for indirect prompt injection. By ingesting user queries and retrieving data from remote Pinecone indexes, the agent could be exposed to malicious instructions embedded within the stored database records.\n
  • Ingestion points: Natural language user queries and data retrieved from vector index records via the search-records tool.\n
  • Boundary markers: The instructions do not specify any delimiters or safety prompts (e.g., 'ignore embedded instructions') for the data being processed.\n
  • Capability inventory: The skill is capable of calling multiple Pinecone MCP tools: search-records, list-indexes, describe-index, describe-index-stats, and rerank-documents.\n
  • Sanitization: No sanitization or validation of the retrieved record metadata is performed before displaying it to the user.\n- [NO_CODE]: The skill consists exclusively of a SKILL.md file containing YAML metadata and markdown instructions. It does not ship with scripts, binaries, or automated installation logic, significantly reducing the attack surface.
Audit Metadata
Risk Level
SAFE
Analyzed
Feb 28, 2026, 01:23 AM