The Agent Skills Directory

[DATA_EXFILTRATION]: The skill is designed to retrieve and process sensitive HR information from ServiceNow. It queries the sn_hr_core_case table for fields such as description, close_notes, and subject_person. These fields frequently contain Personally Identifiable Information (PII) and confidential HR data that are ingested into the AI agent's context without automated filtering.
[COMMAND_EXECUTION]: The skill makes extensive use of the SN-Execute-Background-Script tool (Steps 1, 6, and 7). This tool enables the execution of arbitrary JavaScript on the ServiceNow server. While used for analysis in this skill, the presence of this high-privilege capability is a significant security concern if the agent's instructions are subverted.
[PROMPT_INJECTION]: The skill has a significant attack surface for indirect prompt injection because it processes untrusted text from HR cases without adequate sanitization or boundary protection.
Ingestion points: Untrusted data enters the context via SN-Query-Table and SN-Execute-Background-Script calls targeting the sn_hr_core_case table (SKILL.md, Step 2, 6, 7).
Boundary markers: There are no explicit delimiters or instructions provided to the agent to distinguish between its operational instructions and the potentially malicious content within the case data.
Capability inventory: The skill possesses the ability to create new records (SN-Create-Record) and execute server-side code (SN-Execute-Background-Script), which could be leveraged by an indirect injection attack.
Sanitization: There is no evidence of programmatic filtering or sanitization applied to the retrieved case data before it is interpreted by the agent.

hrsd-kb-generation