cuda-interop

Pass

Audited by Gen Agent Trust Hub on May 23, 2026

Risk Level: SAFENO_CODE
Full Analysis
  • [SAFE]: The skill contains no executable scripts or binary files, relying entirely on instructional content and documentation. It uses basic repository exploration tools (Read, Grep) to assist developers in finding and understanding GPU interop code patterns.
  • [SAFE]: The skill's instructions focus on legitimate GPU programming APIs (CUDA, Vulkan, DLPack) and emphasize correct synchronization practices (events, streams, semaphores) to prevent race conditions.
  • [SAFE]: No network exfiltration, sensitive file access, or obfuscation techniques were detected. The skill references local repository paths for code snippets, which is standard for documentation-centric AI skills.
Audit Metadata
Risk Level
SAFE
Analyzed
May 23, 2026, 03:11 AM
Security Audit — agent-trust-hub — cuda-interop