The Agent Skills Directory

[SAFE]: The skill documents and provides implementation patterns for ServiceNow's official Predictive Intelligence (PI) features, including Classification, Similarity, and Clustering predictors. The code uses standard GlideRecord operations and official sn_ml namespace objects.
[COMMAND_EXECUTION]: The skill includes instructions for using the snow_execute_script tool to test predictions and run server-side JavaScript. This is standard developer functionality for ServiceNow automation and does not involve unauthorized code execution from external sources.
[PROMPT_INJECTION]: The skill facilitates the processing of user-generated content from incident records (e.g., short descriptions) for machine learning predictions. While this introduces a surface for indirect prompt injection, it is treated as a standard data processing task within the ServiceNow environment.
Ingestion points: Untrusted data enters the context via the incident table fields (short_description, description) which are used as input fields for PI classification and similarity models.
Boundary markers: None identified; the scripts pass the record objects directly to the predictor APIs.
Capability inventory: The skill utilizes snow_execute_script for logic execution and snow_query_table for data retrieval.
Sanitization: The skill assumes the ServiceNow machine learning engine and platform-level business rules handle data validation.

predictive-intelligence