tavily-search
Pass
Audited by Gen Agent Trust Hub on Nov 17, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from external websites through search results, extracted page content, and site crawls, creating a vulnerability where embedded instructions could influence the agent.
- Ingestion points: External data enters the context via
tavily_search,tavily_extract, andtavily_crawltools, and via reading saved JSON files in the.tavily/directory. - Boundary markers: The instructions do not prescribe the use of delimiters or 'ignore instructions' warnings when the agent processes retrieved web content.
- Capability inventory: The skill leverages
Bash(tvly *)and suggests the use ofpython3,head, andgrepfor result analysis, forming a chain where untrusted data is processed by shell tools. - Sanitization: No sanitization or filtering of the retrieved web content is performed before it is presented to the agent.
- [EXTERNAL_DOWNLOADS]: The skill directs the installation of the
@tavily/clipackage through global package managers and refers to thetavily-researchskill. These resources are provided by the author, tavily-ai. - [COMMAND_EXECUTION]: The skill uses
tvlyCLI commands for its core functionality and provides Python snippets for parsing locally saved JSON results. This behavior is consistent with the skill's intended purpose.
Audit Metadata