company-research

Pass

Audited by Gen Agent Trust Hub on Oct 22, 2026

Risk Level: SAFE
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from external web searches, which is a common attack surface for indirect prompt injection. \n
  • Ingestion points: Data enters the agent via agent_run and web_search_advanced_exa tool outputs as described in SKILL.md. \n
  • Boundary markers: The skill explicitly instructs on "Token Isolation," advising the use of Task agents and structured output to prevent raw search results from polluting the main context. \n
  • Capability inventory: The skill's primary capabilities are limited to web searching and data synthesis. \n
  • Sanitization: The use of outputSchema ensures that data is processed in a structured format (JSON), which is more resilient to injection than raw natural language. \n- [EXTERNAL_DOWNLOADS]: The skill provides links to external documentation. \n
  • Evidence: References documentation at https://exa.ai/docs/agent/quickstart and other exa.ai subpaths. \n
  • Context: These are official documentation links for the vendor's own search service.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 22, 2026, 07:27 PM