nature-reviewer
Pass
Audited by Gen Agent Trust Hub on Oct 5, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill is designed for scientific manuscript evaluation and does not contain any malicious code, command execution, or unauthorized network operations. Its behavior is consistent with its stated purpose of providing peer-review simulations.
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external content such as manuscripts, abstracts, and author notes. This ingestion surface is managed through highly structured output templates and explicit assessment boundaries, which help mitigate risks associated with processing third-party data. The skill does not request or use high-privilege tools (e.g., network access or sensitive file writes) to process this data.
- [SAFE]: The skill utilizes a relative path to access a shared configuration file (
../nature-shared/core/consistency-sweep.md). This facilitates modularity within the vendor's repository and does not constitute an unsafe file system operation.
Audit Metadata