tavily-search
Pass
Audited by Gen Agent Trust Hub on Sep 30, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill acts as a bridge for the agent to ingest untrusted data from the public internet, creating an attack surface for indirect prompt injection.
- Ingestion points: Results from
tavily_searchand the full-text content retrieved viatavily_extractare directly incorporated into the agent's context in SKILL.md. - Boundary markers: The skill does not provide instructions to wrap or delimit search results with safety markers that would prevent the agent from executing instructions hidden in web content.
- Capability inventory: The agent is granted tools to perform web research and deep extraction of arbitrary URLs.
- Sanitization: There is no mention of filtering, sanitizing, or validating the content retrieved from external sources before it is processed.
Audit Metadata