approval-summarization

Warn

Audited by Gen Agent Trust Hub on Apr 29, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill dynamically assembles and executes JavaScript (GlideScript) on the ServiceNow instance through the SN-Execute-Background-Script tool. This pattern is used in several steps (3, 5, and 7) to extract variables and perform data analysis, representing runtime script generation and execution.
  • [PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection due to its handling of untrusted data from the ServiceNow platform.
  • Ingestion points: External data is ingested from fields like sc_req_item.description and catalog variables via sc_item_option.value as seen in SKILL.md.
  • Boundary markers: The generated summaries (Step 6) do not use clear delimiters or instructions to prevent the agent from obeying instructions embedded within the request items.
  • Capability inventory: The agent environment includes tools for executing scripts (SN-Execute-Background-Script), querying data (SN-Query-Table), and writing notes (SN-Add-Work-Notes).
  • Sanitization: The skill instructions do not specify any validation or sanitization steps for the data retrieved from ServiceNow records before processing.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Apr 29, 2026, 04:36 PM