The Agent Skills Directory

[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection due to its handling of untrusted user-generated content from HR cases.
Ingestion points: Data is ingested from the sys_journal_field table (work notes and additional comments) and sn_hr_core_task (task descriptions/notes) via the SN-Query-Table tool.
Boundary markers: The instructions do not define clear boundaries or provide explicit warnings to the agent to ignore instructions embedded within the retrieved journal entries or task descriptions.
Capability inventory: The skill possesses capabilities to modify ServiceNow data using SN-Update-Record and SN-Add-Work-Notes, and can execute server-side logic via SN-Execute-Background-Script.
Sanitization: While the background script contains a regex to strip HTML tags (.replace(/<[^>]*>/g, '')), this is intended for formatting and does not provide security sanitization against prompt injection attacks.
[COMMAND_EXECUTION]: The skill makes extensive use of the SN-Execute-Background-Script tool, which allows for the execution of arbitrary server-side JavaScript (GlideScript) within the ServiceNow environment.
Evidence: Step 5 and Example 3 use this tool to perform data aggregation and formatting tasks.
Context: While this is a standard administrative capability in the ServiceNow platform and is listed in the skill's allowed tools, it represents a high-privilege execution environment that could be exploited if the agent's logic is subverted through malicious input in the HR case records.

hrsd-sidebar-summarization