ui-actions-policies
Warn
Audited by Gen Agent Trust Hub on Sep 6, 2026
Risk Level: MEDIUMDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [DYNAMIC_EXECUTION]: The skill provides templates and instructions for programmatically creating ServiceNow UI Actions and Policies by generating JavaScript code as strings and inserting them into the sys_ui_action table. Evidence: Section 'Create UI Action Programmatically' in SKILL.md shows uiAction.setValue('script', ...) where the script is a concatenated string.- [INDIRECT_PROMPT_INJECTION]: The skill operates on data ingested from ServiceNow tables and forms, providing a surface where malicious data in those records could potentially influence agent behavior during UI action creation or form analysis. Evidence Chain: 1. Ingestion points: snow_analyze_form tool reads active policies and scripts from ServiceNow; script templates use g_form.getValue and current object to read record data. 2. Boundary markers: Absent in the templates. 3. Capability inventory: snow_create_ui_action and snow_create_ui_policy tools write to the platform; script templates use current.update() and gsftSubmit(). 4. Sanitization: Basic logic checks are present in examples, but no specific sanitization of external data is defined before interpolation into platform logic.
Audit Metadata