deep-research
Pass
Audited by Gen Agent Trust Hub on Oct 29, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill communicates with Google's official Gemini API (generativelanguage.googleapis.com) to perform research tasks. This is a well-known service and the communication is essential for the skill's primary research functionality.
- [INDIRECT_PROMPT_INJECTION]: The skill processes research data retrieved from external sources via the Gemini API, representing a potential surface for indirect prompt injection.
- Ingestion points: Research results are ingested and parsed in scripts/research.py from the provider's JSON response.
- Boundary markers: The script identifies distinct content parts in the API response but does not wrap the final aggregated research report in explicit security delimiters.
- Capability inventory: The script's capabilities are restricted to network requests to the Gemini API and local history management. It lacks dangerous primitives such as arbitrary command execution or system-level file modifications.
- Sanitization: The tool performs structured parsing of the API response and implements size limits to prevent resource exhaustion attacks.
Audit Metadata