The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the snow_execute_script tool to run JavaScript directly on the ServiceNow platform, which is a standard capability for administrative and development tasks.
[DATA_EXPOSURE]: The skill documentation references a local audit log path at ~/.serac/audit/impersonations.jsonl. This path is associated with the vendor's own infrastructure (serac-labs) and is used for auditing impersonation actions performed by the agent.
[INDIRECT_PROMPT_INJECTION]: The skill processes external data from ServiceNow tables and ACL definitions, creating a potential attack surface.
Ingestion points: snow_query_table and snow_acl_explain retrieve data from ServiceNow sys_db_object, sys_security_acl, and user role tables.
Capability inventory: snow_execute_script (arbitrary script execution), snow_record_manage (table write/update operations), and snow_create_acl (security policy creation).
Boundary markers: No specific delimiters or warnings for the agent to ignore instructions embedded within the ServiceNow table data are present.
Sanitization: The instructions do not specify validation or sanitization routines for data retrieved from the ServiceNow instance before processing.
[DYNAMIC_EXECUTION]: The skill demonstrates updating the script and condition columns of ServiceNow sys_security_acl records using the snow_record_manage tool, which is an intended use case for managing dynamic security logic.

acl-security