tavily-remote-mcp

Pass

Audited by Gen Agent Trust Hub on Mar 30, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCREDENTIALS_UNSAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill requires installing '@aiagenta2z/onekey-gateway' and 'ai-agent-marketplace', which are packages provided by the skill author for API access.
  • [CREDENTIALS_UNSAFE]: The Python scripts contain a hardcoded fallback API key 'BETA_TEST_KEY_MARCH_2026' for demo and testing purposes when the environment variable 'DEEPNLP_ONEKEY_ROUTER_ACCESS' is not set.
  • [COMMAND_EXECUTION]: Skill usage documentation instructs the execution of tool calls via 'npx onekey agent', interfacing with the author's gateway CLI tool.
  • [PROMPT_INJECTION]: The skill retrieves and processes content from external websites, which creates an indirect prompt injection surface.
  • Ingestion points: External data enters the agent context via 'tavily_search', 'tavily_crawl', and 'tavily_extract' tools.
  • Boundary markers: No delimiters or safety instructions are used in the provided scripts to wrap the fetched content.
  • Capability inventory: The skill facilitates network operations to fetch web data and present it back to the agent.
  • Sanitization: No validation or filtering of the retrieved web content is performed.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 30, 2026, 02:41 PM
Security Audit — agent-trust-hub — tavily-remote-mcp