government-case-summarization

Pass

Audited by Gen Agent Trust Hub on May 20, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection (Category 8) as it ingests and summarizes untrusted text data from external sources.
  • Ingestion points: Data is retrieved from ServiceNow tables in SKILL.md, specifically the description field from sn_psds_case (Step 1) and the value field from sys_journal_field (Step 7).
  • Boundary markers: The summary template provided in Step 8 lacks explicit delimiters or instructions for the AI to ignore potentially malicious content embedded within the retrieved data variables (e.g., `[description
  • summarized], [activity summary]`).
  • Capability inventory: The skill primarily performs data retrieval and text generation. While it does not invoke dangerous system-level tools based on the summarized content, a successful injection could influence the agent's tone or subsequent interactions.
  • Sanitization: There are no instructions for the agent to sanitize or validate the content of the journal fields or case descriptions before they are interpolated into the final summary.
Audit Metadata
Risk Level
SAFE
Analyzed
May 20, 2026, 01:34 PM
Security Audit — agent-trust-hub — government-case-summarization