tavily-search

Pass

Audited by Gen Agent Trust Hub on Jul 21, 2026

Risk Level: SAFE
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill references an installation script located at https://cli.tavily.com/install.sh. This is an official resource provided by the vendor (tavily-ai) for setting up their search utility.
  • [REMOTE_CODE_EXECUTION]: The documentation includes a setup command that pipes a remote script to the shell (curl | bash). This is the vendor's recommended installation method for the required CLI tool and is considered a normal operational step for this service.
  • [COMMAND_EXECUTION]: The skill uses the tvly command to interact with the search API. Execution is scoped to this specific tool through the platform's tool restrictions.
  • [PROMPT_INJECTION]: The skill acts as an interface for web search results, which constitutes an ingestion point for untrusted external data.
  • Ingestion points: Search results from tvly search (SKILL.md).
  • Boundary markers: The instructions encourage using the --json flag to ensure data is returned in a structured format.
  • Capability inventory: The skill is limited to the tvly command for search and login operations.
  • Sanitization: Results are returned as structured data snippets optimized for model consumption, reducing the risk of unintended instruction execution.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 21, 2026, 12:09 AM
Security Audit — agent-trust-hub — tavily-search