gemini-deep-research
Pass
Audited by Gen Agent Trust Hub on Sep 21, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied research queries and custom formatting instructions which are directly interpolated into the API request payload without boundary markers or isolation techniques. This creates a surface where a malicious query could attempt to override the underlying AI agent's behavior.
- Ingestion points: The
--queryand--formatparameters inscripts/deep_research.py. - Boundary markers: None present; the query is appended directly to the instructions.
- Capability inventory: The skill performs authenticated network requests and writes generated reports to the local file system.
- Sanitization: No validation or escaping is applied to the input strings.
- [EXTERNAL_DOWNLOADS]: The skill's research script communicates with Google's generative language API at
https://generativelanguage.googleapis.com. This is a well-known service required for the skill's primary functionality.
Audit Metadata