exa-web-search
Pass
Audited by Gen Agent Trust Hub on Aug 3, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill ingests untrusted search results from the web via the
web_search_exatool, creating a surface for indirect prompt injection. Malicious instructions embedded in searched content could potentially influence the agent's behavior. - Ingestion points: Web content retrieved by the
web_search_exatool as described inSKILL.md. - Boundary markers: No delimiters are specified to distinguish between retrieved content and system instructions.
- Capability inventory: The skill is restricted to the
mcp(exa:web_search_exa)tool. - Sanitization: The instructions do not define methods for sanitizing or escaping content retrieved from external sources.
Audit Metadata