snowflake-docs

Pass

Audited by Gen Agent Trust Hub on Aug 14, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes SQL statements using the sql_execute tool to manage the documentation database and perform search queries. The interpolation of user questions into the JSON search query string in SKILL.md represents a surface for query manipulation if input is not properly escaped.
  • [EXTERNAL_DOWNLOADS]: The skill programmatically requests the installation of an official Snowflake Marketplace listing (ID GZSTZ67BY9OQ4) to enable search functionality. This is a legitimate setup step for this vendor-provided utility.
  • [INDIRECT_PROMPT_INJECTION]: The skill retrieves documentation fragments from an external search service and incorporates them into the agent's context.
  • Ingestion points: Document chunks retrieved from the SNOWFLAKE.CORTEX.SEARCH_PREVIEW function as specified in SKILL.md.
  • Boundary markers: None used; the skill lacks explicit delimiters to isolate retrieved content from primary instructions.
  • Capability inventory: The agent has access to sql_execute for administrative setup and data retrieval.
  • Sanitization: No explicit sanitization is performed on the retrieved documentation content, as it originates from a trusted official source.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 14, 2026, 11:02 PM
Security Audit — agent-trust-hub — snowflake-docs