enterprise-context

Pass

Audited by Gen Agent Trust Hub on Jun 17, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill provides instructions aimed at improving the reliability of agent responses through systematic searching and source attribution.\n- [PROMPT_INJECTION]: The skill describes an indirect prompt injection attack surface.\n
  • Ingestion points: The skill ingests data from gemini_enterprise_search and gemini_enterprise_get_document (SKILL.md).\n
  • Boundary markers: Absent; there are no instructions to use delimiters or ignore instructions within the search results.\n
  • Capability inventory: The agent can retrieve and read full documents based on search queries (SKILL.md).\n
  • Sanitization: Absent; the skill does not require validation or filtering of document content before processing.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 17, 2026, 12:21 PM
Security Audit — agent-trust-hub — enterprise-context