The Agent Skills Directory

[PROMPT_INJECTION] (LOW): The skill presents an indirect prompt injection surface where untrusted input could influence SQL execution within the Databricks environment.
Ingestion points: The manage_metric_views tool accepts free-text SQL expressions in fields such as dimensions, measures, filter_expr, and where.
Boundary markers: No specific delimiters or validation logic is documented to separate user data from the SQL command structure.
Capability inventory: The skill can execute SQL to create, alter, and query database views, and modify permissions via the grant action.
Sanitization: Documentation does not indicate any sanitization or escaping of the provided SQL strings before they are incorporated into the YAML definition or executed.
[COMMAND_EXECUTION] (LOW): The skill provides an interface for administrative database operations.
The manage_metric_views tool includes actions for grant, drop, and alter, which can modify security policies and data structures in a production environment.
While these are intended functionalities, they necessitate strict access control to prevent unauthorized data modification or privilege escalation.

metric-views