The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from multiple ServiceNow compliance tables which could serve as an attack surface for indirect prompt injection.
Ingestion points: Data is retrieved from sn_compliance_control_objective, sn_compliance_control, sn_compliance_policy, and sn_compliance_citation via both the SN-Query-Table tool and REST API calls.
Boundary markers: The instructions lack explicit boundary markers or delimiters to instruct the agent to ignore potentially malicious content embedded within the compliance records.
Capability inventory: The skill possesses significant capabilities including the ability to create records (SN-Create-Record), update records (SN-Update-Record), and execute server-side scripts (SN-Execute-Background-Script).
Sanitization: There is no evidence of sanitization or validation of the content retrieved from the ServiceNow instance before it is processed by the agent.
[COMMAND_EXECUTION]: The skill uses the Bash tool to perform network operations.
Evidence: Multiple steps (Step 1, 2, 6, 7) provide curl commands to interact with the ServiceNow Table API. These commands are directed at local ServiceNow endpoints (e.g., /api/now/table/sn_compliance_control_objective).
[DYNAMIC_EXECUTION]: The skill leverages the SN-Execute-Background-Script tool to perform complex analysis.
Evidence: Step 4 and Step 5 contain server-side JavaScript intended for execution on the ServiceNow instance to calculate effectiveness metrics and identify compliance gaps. While the scripts themselves are statically defined in the skill documentation, they represent a powerful execution vector within the target environment.

control-objective-management