The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it retrieves and analyzes data from ServiceNow records (incidents, search logs, and user feedback) which can be populated with malicious instructions by external actors.
Ingestion points: Data is ingested from the incident, kb_submission, search_log, and kb_feedback tables using the SN-Query-Table and SN-Aggregate tools as described in Steps 1 through 4.
Boundary markers: Absent. The skill instructions do not provide delimiters or clear guidance to the agent to distinguish between its core instructions and the external data it is processing.
Capability inventory: The skill possesses write capabilities through the SN-Create-Record and SN-Update-Record tools, allowing it to commit the results of its analysis back to the ServiceNow instance. It also lists Bash as an available native tool.
Sanitization: Absent. There is no evidence of input validation or sanitization of the data retrieved from ServiceNow before it is used to generate summaries or action plans.

gap-grouping