web-search-tavily
Pass
Audited by Gen Agent Trust Hub on Sep 17, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: As a web search utility, this skill retrieves content from external websites which could potentially contain malicious instructions intended to mislead or exploit the agent.
- Ingestion points: Data enters the system via the
tavilySearchfunction inscripts/search.ts, specifically thecontent,raw_content, andanswerfields returned by the Tavily API. - Boundary markers: The output uses basic visual formatting (horizontal lines and numbering) but lacks specific instructions to the agent to treat the following data as untrusted or to ignore embedded commands.
- Capability inventory: The script has permissions to read environment variables (
TAVILY_API_KEY) and make network requests toapi.tavily.com. - Sanitization: No filtering or sanitization of the retrieved web content is performed by the script before it is output for agent consumption.
Audit Metadata