exa-deep-search
Pass
Audited by Gen Agent Trust Hub on Oct 4, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes external content from the web, creating a surface for indirect prompt injection attacks.
- Ingestion points: Untrusted data enters the agent context through the
exa_search(highlights, summaries) andexa_contents(full page text) tools as defined inSKILL.md. - Boundary markers: The instructions lack explicit requirements for using secure delimiters or wrappers to isolate untrusted external content from the agent's core instructions.
- Capability inventory: The skill uses the
sandbase_call_toolgateway, which allows for dynamic tool execution and network-based search operations. - Sanitization: No specific sanitization or filtering of the extracted web content is mentioned to prevent embedded instructions from influencing agent behavior.
Audit Metadata