The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the agent to clone the op-plugin repository from https://gitcode.com/ascend/op-plugin.git. This is an official source for Ascend PyTorch integration tools and aligns with the skill's purpose.- [COMMAND_EXECUTION]: The instructions involve standard development and build commands, including python setup.py bdist_wheel, pip install, and cmake. It also provides templates for executing commands inside Docker or Podman containers via docker exec to support isolated build environments, which is common in NPU development.- [REMOTE_CODE_EXECUTION]: The documentation references torch.utils.cpp_extension.load for dynamically compiling and loading C++ extensions. While this executes generated code, it is the standard and expected mechanism for PyTorch operator development and is used here in a legitimate context.- [DATA_EXFILTRATION]: No patterns of unauthorized data access or external transmission of sensitive information were found. Network operations are limited to cloning official repositories and installing standard development packages.- [PROMPT_INJECTION]: The skill contains behavioral guidelines for the agent (under section 0. 重要原则) to ensure it follows best practices for operator integration, but no malicious override, safety bypass, or system prompt extraction attempts were identified.

ascend-opplugin