write-script-snowflake
Pass
Audited by Gen Agent Trust Hub on Oct 13, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to use
wmillCLI commands (wmill script preview,wmill script run,wmill generate-metadata,wmill sync push) for script management and execution. These tools are official vendor resources from windmill-labs used to interact with the platform. - [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data via S3 objects and script arguments. While it mandates the use of
?placeholders for SQL parameterization to mitigate SQL injection, the ingestion of external content remains a surface for indirect prompt injection. - Ingestion points: S3 files (via
s3objecttype) and user-supplied arguments (via the-dflag inwmill script preview). - Boundary markers: The instructions enforce the use of Snowflake
?placeholders to separate SQL code from external data. - Capability inventory: The agent can generate SQL code, write it to local files, and execute it using the
wmillCLI. - Sanitization: SQL parameterization is explicitly required.
- [DYNAMIC_EXECUTION]: The skill is designed to generate and execute Snowflake SQL scripts at runtime based on user requirements. This involves creating executable queries and invoking them via the
wmill script previewcommand.
Audit Metadata