accel-pytorch-code-porting
Pass
Audited by Gen Agent Trust Hub on Sep 27, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill contains legitimate technical documentation, code templates, and migration strategies for hardware acceleration. It promotes best practices like backend-neutral abstractions and explicit verification steps.
- [EXTERNAL_DOWNLOADS]: References official documentation and code repositories for Moore Threads (torch_musa, torchada) on GitHub. These are well-known hardware vendor resources used for providing necessary context and drivers for the porting task.
- [COMMAND_EXECUTION]: Includes standard diagnostic shell commands such as
mthreads-gmi,musaInfo, andmcc --version. These are routine system checks used to verify the health and configuration of the local accelerator environment before code execution. - [INDIRECT_PROMPT_INJECTION]: The skill contains logic for processing repository content. However, it includes explicit instructions for the agent to respect local repository safety guidelines (
AGENTS.md) and environment documentation, reducing the risk of unintended behavior from untrusted data.
Audit Metadata