web-search

Pass

Audited by Gen Agent Trust Hub on Aug 31, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests data from external websites via the webFetch and webSearch functions in exa_tool.mjs. This represents a potential attack surface for indirect prompt injection where external content could contain instructions designed to manipulate the agent.\n
  • Ingestion points: External data is fetched from arbitrary URLs provided to the tool.\n
  • Boundary markers: The output format uses simple text blocks and does not include specific delimiters or warnings to the agent regarding untrusted content.\n
  • Capability inventory: The script is limited to network requests via fetch and does not possess capabilities for file modification or arbitrary command execution.\n
  • Sanitization: Results are sanitized by removing URL tracking parameters and truncating titles and excerpts to predefined limits.\n- [EXTERNAL_DOWNLOADS]: The script communicates with api.exa.ai and mcp.exa.ai to retrieve search data. These requests are directed to the official service endpoints required for the tool's intended functionality and do not involve untrusted third-party domains.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 31, 2026, 11:53 AM
Security Audit — agent-trust-hub — web-search