snowflake-development
Warn
Audited by Gen Agent Trust Hub on Jun 16, 2026
Risk Level: MEDIUMPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The script
scripts/snowflake_query_helper.pyis vulnerable to indirect prompt injection through SQL injection in generated templates. - Ingestion points: Command-line arguments for the
merge,dynamic-table, andgrantsubcommands (e.g.,--target,--role,--columns,--key) inscripts/snowflake_query_helper.py. - Boundary markers: None. The script uses f-strings and
textwrap.dedentto assemble SQL statements without escaping or parameterizing the input values. - Capability inventory: The agent is instructed in
SKILL.mdto execute this script to generate MERGE statements, Dynamic Table DDL, and RBAC grants. - Sanitization: Absent. Input strings are used directly in SQL identifiers and logic, allowing a malicious user to supply object names that terminate the intended statement and execute arbitrary SQL commands if the generated output is subsequently run in a Snowflake environment.
Audit Metadata