The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the SN-Execute-Background-Script tool to perform server-side processing within the ServiceNow environment.
Evidence: Found in Step 6 and Example 3 of SKILL.md, where arbitrary JavaScript (GlideScript) is executed to perform confidentiality checks and generate email templates.
Risk: There is a potential for script injection if untrusted case data or email content is interpolated into the script string without proper escaping.
[DATA_EXFILTRATION]: The skill is designed to retrieve and process sensitive Human Resources data, including Personally Identifiable Information (PII).
Evidence: Data retrieval operations in Step 1 and Step 2 of SKILL.md access tables such as sn_hr_core_case and sys_email, specifically targeting fields like subject_person, description, and body_text.
[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface by ingesting untrusted data from external users (employees) which is then used to influence agent output.
Ingestion points: sn_hr_core_case (description, comments) and sys_email (body_text) as documented in the retrieval steps of SKILL.md.
Boundary markers: Absent; there are no specific delimiters or instructions to ignore embedded commands within the ingested text.
Capability inventory: The agent has access to SN-Update-Record, SN-Add-Work-Notes, and SN-Execute-Background-Script tools.
Sanitization: While Step 6 includes a regex-based PII scanner, there is no validation or sanitization implemented to detect or block malicious instructions embedded in the case data.

hrsd-email-recommendation