snowflake-development
Pass
Audited by Gen Agent Trust Hub on Oct 26, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill provides templates for using Snowflake Cortex AI functions and Agents which process potentially untrusted external data.
- Ingestion points: Data is ingested from Snowflake tables (e.g.,
ticket_text) and stage files (e.g.,model.yaml) into LLM contexts through functions likeAI_COMPLETE,AI_CLASSIFY, and Cortex Agents. - Boundary markers: The skill includes a 'Limitations' section instructing the agent to stop and ask for clarification if safety boundaries or success criteria are missing.
- Capability inventory: The skill utilizes SQL DDL/DML execution, Snowpark Python environments, and AI completion/classification tools.
- Sanitization: The instructions explicitly recommend security hardening measures, including masking policies for PII columns, row access policies, and least-privilege RBAC.
- [SAFE]: The skill follows security best practices by instructing users to use environment variables for authentication instead of hardcoding credentials in Snowpark Python scripts.
Audit Metadata