scientific-deep-research
Pass
Audited by Gen Agent Trust Hub on Jul 15, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill is designed for rigorous scientific research, utilizing a 'Think-Report-Action' cycle adapted for academic contexts.
- [EXTERNAL_DOWNLOADS]: The skill interacts with well-known and trusted academic databases, including PubMed, arXiv, Google Scholar, and Semantic Scholar, to retrieve scientific literature. These are recognized services for the intended research purpose.
- [COMMAND_EXECUTION]: Uses established research tools from the ToolUniverse SMCP to perform literature searches and metadata lookups. All tool usage is consistent with the skill's stated goal of literature review.
- [PROMPT_INJECTION]: No malicious prompt injection patterns were found. The instructions prioritize hallucination prevention and source tracking, reinforcing the agent's adherence to factual research rather than bypassing safety guidelines.
- [DATA_EXFILTRATION]: No evidence of sensitive file access (such as credentials or environment variables) or unauthorized network exfiltration was detected. Network operations are limited to academic search domains.
- [INDIRECT_PROMPT_INJECTION]: While the skill ingests external research data which represents a theoretical attack surface, it implements mitigating procedures such as evidence hierarchy scoring (GRADE approach), critical appraisal checklists (Cochrane, Newcastle-Ottawa), and cross-validation across multiple independent sources.
Audit Metadata