nature-statistics
Pass
Audited by Gen Agent Trust Hub on Aug 23, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill is entirely focused on scientific manuscript auditing. It provides instructional checklists for identifying statistical risks such as pseudoreplication, multiple comparison errors, and reporting gaps.
- [DATA_EXPOSURE_AND_EXFILTRATION]: The skill references established scientific domains (nature.com) for official guidelines. It does not access sensitive system paths, hardcode credentials, or perform network requests to untrusted servers.
- [COMMAND_EXECUTION]: While the workflow instructions mention generating code for reanalysis if requested by the user, the skill does not include any automated shell command execution or remote script invocation patterns.
- [PROMPT_INJECTION]: The instructions do not contain attempts to bypass AI safety filters or override core agent behavior. It maintains a strict boundary against inventing data or p-values, prioritizing user-confirmed facts.
Audit Metadata