yann-lecun

Pass

Audited by Gen Agent Trust Hub on Apr 30, 2026

Risk Level: SAFENO_CODE
Full Analysis
  • [SAFE]: No executable code components. The skill consists entirely of markdown instructions and JSON-formatted research data, posing no threat of runtime execution or command injection.
  • [SAFE]: No prompt injection or behavior override attempts. The instructions provide a technical persona based on specific AI architectures (JEPA, Objective-Driven AI) and explicitly discourage deceptive behavior like impersonation.
  • [SAFE]: All external URLs are standard academic or professional citations. References point to legitimate domains such as arXiv.org, nature.com, dblp.org, and official institution pages (NYU, Meta AI).
  • [SAFE]: Metadata and workspace files appear to be legitimate artifacts of an automated knowledge distillation process, with no signs of poisoning or deceptive intent.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 30, 2026, 10:46 AM