The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses SN-Execute-Background-Script to execute arbitrary JavaScript on the ServiceNow server. This is a high-privilege administrative tool that allows the agent to run code with server-side permissions.
[DATA_EXFILTRATION]: Scans for and exposes sensitive data categories. Section 1.1 identifies fields containing passwords, tokens, and API keys. Section 1.3 samples actual data from sensitive fields (like email and phone) and logs the results to the console. Although it attempts masking, it still provides the agent with access to sensitive record values.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from various ServiceNow tables to identify embedded PII. This creates a surface where malicious instructions in record content could be ingested by the agent.
Ingestion points: Scans incident.description, incident.comments, sc_req_item.special_instructions, and kb_knowledge.text across the instance.
Boundary markers: None. The scripts process raw string content without using delimiters or instructions to ignore embedded commands.
Capability inventory: High; includes table queries, record creation/modification, and background script execution.
Sanitization: Absent. Data is read, matched via regex, and logged directly to the output.
[DYNAMIC_EXECUTION]: Generates and installs persistent server-side code. Section 2.3 uses SN-Create-Record to inject scripts into the sys_script table (Business Rules), which are then executed by the system during standard user interactions.

data-classification