The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes data from external ServiceNow tables (such as sc_req_item, sc_item_option_mtom, and sysapproval_approver) which could contain untrusted user-supplied content.
Ingestion points: The getRITMVariables function reads variable values from the sc_item_option_mtom table, and createServiceRequest accepts a variables object and catalogItemName string from the agent's context (SKILL.md).
Boundary markers: There are no explicit boundary markers or instructions to ignore embedded commands when processing these strings.
Capability inventory: The skill includes the snow_execute_script tool, which allows for arbitrary server-side JavaScript execution on the ServiceNow instance, and snow_query_table for data retrieval.
Sanitization: There is no evidence of sanitization or escaping of the values retrieved from ServiceNow records before they are processed or logged.
[DYNAMIC_EXECUTION]: The skill provides and documents the use of the snow_execute_script tool, which serves as a mechanism for executing dynamic code at runtime.
Evidence: The 'MCP Tool Integration' section explicitly demonstrates wrapping ServiceNow JavaScript (ES5) within a string and passing it to the snow_execute_script tool for execution on the target platform.

request-management