exa-web-search

Pass

Audited by Gen Agent Trust Hub on Aug 10, 2026

Risk Level: SAFE
Full Analysis
  • [DATA_EXFILTRATION]: The skill performs network requests to api.exa.ai, which is the official endpoint for the well-known Exa AI search service. It retrieves the required EXA_API_KEY from environment variables or local .env files using standard initialization patterns. No sensitive local data is accessed or transmitted beyond the API key required for the service.
  • [EXTERNAL_DOWNLOADS]: The skill documentation describes standard installation via repository cloning. The scripts themselves do not download or execute external code at runtime and have no external dependencies, relying entirely on the Python standard library.
  • [PROMPT_INJECTION]: The skill processes untrusted content from the web, which constitutes an indirect prompt injection surface.
  • Ingestion points: The search and contents subcommands in scripts/exa_client.py retrieve snippets and full text from external web pages.
  • Boundary markers: The results are formatted in Markdown with clear headers and list structures to help the agent distinguish between search results and instructions.
  • Capability inventory: The skill is restricted to web retrieval and does not contain dangerous functions such as eval(), exec(), or arbitrary shell command execution.
  • Sanitization: Data retrieved from the API is presented to the agent context without additional filtering or sanitization, which is typical for search-oriented tools.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 10, 2026, 04:10 PM
Security Audit — agent-trust-hub — exa-web-search