The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes several shell commands to validate the execution environment, including checking NPU status (npu-smi) and verifying torch_npu availability. It also executes the final generated profiling scripts using bash and python.
[REMOTE_CODE_EXECUTION]: The skill employs dynamic script generation and execution. It reads user-provided scripts, creates a modified copy containing torch_npu.profiler instrumentation code, and executes these scripts locally to collect performance metrics.
[DATA_EXFILTRATION]: The skill accesses local file system paths and reads the contents of scripts provided by the user. This access is necessary for its stated purpose of performance analysis, and no evidence of data transmission to external domains was detected.

pytorch-profiling-collection