The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the Snowflake CLI tool (snow) to execute SQL commands, validate DDL statements, and perform database operations. This includes running snow sql with user-provided or agent-generated queries.
[INDIRECT_PROMPT_INJECTION]: The skill is instructed to read Snowflake table, view, and column comments as a primary source for generating synonyms and documentation in new semantic views. This behavior creates a vulnerability surface where an attacker with control over database metadata could inject instructions to influence the agent's DDL generation or query execution.
Ingestion points: Snowflake table/view/column comments are read via SQL statements as described in SKILL.md (Workflow Step 4 and Synonyms section).
Boundary markers: No explicit delimiters or instructions to ignore embedded commands within the fetched comments are defined.
Capability inventory: The skill has the ability to execute arbitrary SQL via snow sql (Workflow Steps 7, 9, and 10).
Sanitization: The instructions do not specify any validation or sanitization of the retrieved metadata before it is interpolated into the DDL drafts.

snowflake-semanticview