rwkv-architecture

Pass

Audited by Gen Agent Trust Hub on Sep 17, 2026

Risk Level: SAFE
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill references the installation of standard machine learning dependencies including torch, rwkv, and transformers. It utilizes the official PyTorch package index at download.pytorch.org for GPU-accelerated versions, which is a trusted and well-known service for machine learning distribution.
  • [DYNAMIC_EXECUTION]: Implementation examples include the use of environment variables such as RWKV_JIT_ON and RWKV_CUDA_ON to enable performance-optimizing Just-In-Time compilation and custom CUDA kernels. This is standard behavior for the rwkv library to achieve efficient inference.
  • [INDIRECT_PROMPT_INJECTION]: The skill documents the capability of the RWKV architecture to process large-scale text sequences from external sources, which is an inherent ingestion surface for language models. The provided examples focus on model forward passes and sequence processing without instructions that would lead to unauthorized command execution or safety bypasses.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 17, 2026, 07:53 PM