deep-research
Pass
Audited by Gen Agent Trust Hub on Mar 10, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill is a comprehensive tool for academic research. All instructions and agent roles (e.g., bibliography_agent, report_compiler_agent) are strictly aligned with this legitimate purpose.
- [EXTERNAL_DOWNLOADS]: The skill references well-known academic platforms and databases such as PubMed, arXiv, Google Scholar, OSF, and PROSPERO. These are well-known services within the research community and are used for their intended purpose.
- [PROMPT_INJECTION]: No evidence was found of instructions attempting to bypass safety filters, extract system prompts, or override agent constraints. The use of a 'Socratic Mentor' persona is a pedagogical technique and not a malicious jailbreak.
- [DATA_EXFILTRATION]: There are no commands or instructions designed to access sensitive local file paths (like SSH keys or environment variables) or to send data to untrusted external domains.
- [INDIRECT_PROMPT_INJECTION]: The skill processes external literature. It mitigates potential risks from untrusted data by incorporating dedicated 'ethics_review_agent' and 'source_verification_agent' components that systematically evaluate the quality and integrity of sources.
- [COMMAND_EXECUTION]: The skill mentions statistical software and packages (e.g., R's metafor, Python's statsmodels) for the user's reference in conducting meta-analysis, but it does not contain code to execute these at runtime on the host system.
Audit Metadata