The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill utilizes the snow_execute_script tool to run server-side JavaScript (ES5) on a ServiceNow instance. This enables runtime execution of code strings for operational tasks.
[INDIRECT_PROMPT_INJECTION]: The skill processes data from records in tables like sn_si_incident, which could contain malicious instructions designed to influence the agent's decision-making or playbook execution.
Ingestion points: Data retrieved from tables including sn_si_incident, sn_vul_vulnerable_item, and sn_ti_indicator via snow_query_table in SKILL.md.
Boundary markers: None identified in SKILL.md to delimit or isolate untrusted record content from the agent's instructional context.
Capability inventory: Access to snow_execute_script, snow_create_event, and the ability to update records (e.g., incident.update(), user.update()) as seen in SKILL.md.
Sanitization: There is no evidence of input validation or sanitization for data retrieved from record fields before it is processed by the playbooks.
[PRIVILEGE_ESCALATION]: The executeContainmentPlaybook function in SKILL.md contains logic to modify user account status (setting active to false and locked_out to true) and initiate infrastructure isolation events. These represent high-privilege administrative actions.

security-operations