deep-research-framework
Pass
Audited by Gen Agent Trust Hub on Oct 17, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to process external information from various sources including official documents, press, and social media, creating a potential surface for indirect prompt injection where instructions could be hidden in external content.
- Ingestion points: External research sources identified through web searches (Primary, Secondary, and Tertiary sources).
- Boundary markers: The workflow requires labeling unconfirmed facts as
[UNVERIFIED]and defines a strict source hierarchy to help differentiate content trust levels. - Capability inventory: The skill contains no executable code or file-writing operations, relying instead on the agent's native web access capabilities.
- Sanitization: The skill enforces a cross-verification requirement (two independent sources for key conclusions) and source tiering to mitigate the impact of untrusted or biased data.
Audit Metadata