web-search-exa

Pass

Audited by Gen Agent Trust Hub on May 19, 2026

Risk Level: SAFENO_CODEEXTERNAL_DOWNLOADS
Full Analysis
  • [SAFE]: The skill consists entirely of Markdown instructions and JSON metadata. No executable scripts, binaries, or active code components are present.
  • [EXTERNAL_DOWNLOADS]: The skill configuration directs the agent to connect to the official Exa MCP server at https://mcp.exa.ai/mcp. Exa is a well-known service provider for neural search and AI-ready content extraction.
  • [PROMPT_INJECTION]: As the skill's primary function is web search and crawling, it inherently introduces a surface for indirect prompt injection. Malicious instructions hosted on third-party websites could be ingested by the agent during a search or crawl operation. This is a known risk for all web-retrieval tools and is managed by the underlying agent platform's safety filters.
  • Ingestion points: Search results and page content retrieved via web_search_exa and crawling_exa in SKILL.md.
  • Boundary markers: None explicitly defined in the skill instructions.
  • Capability inventory: Network access to the Exa API for search and retrieval.
  • Sanitization: None described within the skill body.
Audit Metadata
Risk Level
SAFE
Analyzed
May 19, 2026, 10:47 AM
Security Audit — agent-trust-hub — web-search-exa