The Agent Skills Directory

[DYNAMIC_EXECUTION]: The troubleshooting section (Common Issue 3) instructs the agent to use eval() within a ServiceNow background script to check the syntax of code retrieved from the database (<script_content>). This pattern executes arbitrary code from the data source at runtime, which is a significant security risk if the source content is untrusted or compromised.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process configuration data from multiple ServiceNow tables, creating a surface for indirect prompt injection.
Ingestion points: Data is ingested from tables such as sys_script, wf_activity, and sc_cat_item via the SN-Query-Table tool and ServiceNow REST APIs (defined in SKILL.md).
Boundary markers: No specific delimiters or boundary markers are defined to isolate the ingested configuration data from the agent's primary instructions.
Capability inventory: The skill utilizes the Bash native tool and the SN-Execute-Background-Script tool (mentioned in Troubleshooting Step 3) which provide broad execution capabilities.
Sanitization: The instructions lack any sanitization or validation logic to filter potentially malicious instructions or code embedded within the ServiceNow records before they are processed or executed.
[COMMAND_EXECUTION]: The skill utilizes the Bash tool for executing validation scripts. While standard for administrative tasks, the broad nature of shell access combined with the processing of external data increases the potential attack surface if inputs are not strictly controlled.

configuration-validation