sota-ml-engineering
Pass
Audited by Gen Agent Trust Hub on Nov 6, 2026
Risk Level: SAFENO_CODE
Full Analysis
- [NO_CODE]: The skill consists entirely of Markdown (.md) files. There are no executable scripts (.py, .js, .sh), configuration files that trigger automation, or remote code downloads.
- [SAFE]: The documentation references well-known and trusted research papers (Google Research), official guides (Google Developers), and established security frameworks (MITRE ATLAS, NIST AI RMF, OWASP). These external links are for instructional reference only.
- [SAFE]: While the skill contains
grepand shell command strings within its 'Audit checklist' sections, these are provided as documentation/examples for a human auditor to run manually. They are not executed by the agent automatically, nor do they contain malicious injection patterns. - [SAFE]: The content specifically instructs on how to defend against security threats such as unsafe pickle deserialization, training data poisoning, and model extraction. It advocates for security-best practices like using
safetensorsand verifying model signatures.
Audit Metadata