tavily-web
Pass
Audited by Gen Agent Trust Hub on Oct 4, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The installation instructions direct the user to fetch and install the skill from a third-party GitHub repository (
BenedictKing/tavily-web) using thenpx skills addcommand. - [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data retrieved from web searches and extracted URL content, creating a risk for indirect prompt injection.
- Ingestion points: Search result summaries and full content extracted from external URLs via the Tavily API (SKILL.md).
- Boundary markers: The documentation does not define specific delimiters or instructional warnings to prevent the agent from obeying commands embedded in retrieved web content.
- Capability inventory: The skill uses the Tavily API to perform web searches, crawl sites, and extract content, which can then be processed by the agent.
- Sanitization: No mention of sanitization, filtering, or validation of the content retrieved from the web before it is analyzed by the agent.
Audit Metadata