The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes natural language process descriptions to generate structured bot actions, triggers, and execution logic. This creates a vulnerability surface where a malicious process description could potentially influence the agent to generate harmful automation steps or bypass intended logic.
Ingestion points: Natural language process descriptions provided by the user in Step 1.
Boundary markers: Absent; there are no instructions to the agent to treat the input descriptions as untrusted data or use specific delimiters.
Capability inventory: SN-Create-Record for creating bot definitions and SN-Execute-Background-Script for running server-side logic.
Sanitization: No input validation or sanitization of the process description is defined.
[COMMAND_EXECUTION]: The skill leverages the SN-Execute-Background-Script MCP tool to run administrative scripts on the ServiceNow instance.
Evidence: Step 7 ('Monitor Bot Execution') provides a JavaScript template to be executed as a background script to generate performance analytics.
[DYNAMIC_EXECUTION]: The skill generates and executes server-side JavaScript code at runtime to perform data aggregation and analysis.
Evidence: The JavaScript code block provided in Step 7 uses GlideAggregate and gs.info to perform server-side processing on ServiceNow execution logs.

rpa-bot-generation