web-search-brave
Pass
Audited by Gen Agent Trust Hub on Sep 17, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill facilitates the retrieval of external data from the web through search results, which represents a standard surface for indirect prompt injection where third-party content could attempt to influence agent behavior.
- Ingestion points: Search result metadata (titles, descriptions, and snippets) are fetched from
api.search.brave.cominscripts/search.ts. - Boundary markers: The script does not wrap search results in specific delimiters or boundary markers before they are presented to the agent.
- Capability inventory: Network access is restricted to the
api.search.brave.comdomain; no other sensitive system capabilities are invoked. - Sanitization: The script extracts and returns only specific, structured fields from the API's JSON response, preventing the injection of raw, unstructured response data.
Audit Metadata