experimental-design
Pass
Audited by Gen Agent Trust Hub on Oct 23, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill requires the installation of standard scientific Python packages including
numpy,pandas, andpydoefrom official registries using theuvpackage manager. - [INDIRECT_PROMPT_INJECTION]: The skill processes natural language descriptions of experiments to suggest designs. It mitigates injection risks by using structured data processing and standard libraries to generate CSV outputs rather than executing commands based on user-provided strings.
- [SAFE]: Technical analysis of the bundled Python scripts (
randomization.pyanddoe_designs.py) confirms they perform strictly computational statistical tasks without accessing the network, sensitive file paths, or system environment. No signs of obfuscation, hardcoded credentials, or malicious metadata were detected.
Audit Metadata