The Agent Skills Directory

[COMMAND_EXECUTION]: The skill guides the agent to use shell commands such as bash, sed, and cp to manage model configurations and execute inference benchmark scripts (infer.sh, runner_*.py). These commands are performed locally and are necessary for the primary purpose of NPU performance tuning and validation.
[DATA_EXPOSURE]: The skill accesses local model definitions and YAML configurations. It does not attempt to access sensitive system files, environment secrets, or credentials. All file operations are scoped to the cann-recipes-infer project directory.
[EXTERNAL_DOWNLOADS]: The skill references official Huawei Ascend documentation (hiascend.com) and a repository on GitCode (gitcode.com/cann), which are established sources for CANN (Compute Architecture for Neural Networks) development resources. These references are documented neutrally as part of the technical requirements.
[SAFE]: The skill implements a structured optimization workflow that includes mandatory performance testing and the generation of an optimization report. No malicious patterns, obfuscation, or unauthorized network activity were detected.

external-cannbot-model-model-infer-superkernel