Nvidia smi tutorial
v1.0.0Provides detailed guidance on using the nvidia-smi command for real-time NVIDIA GPU monitoring, management, troubleshooting, and automation.
⭐ 0· 102·0 current·0 all-time
bywow@duanc-chao
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
OpenClaw
Benign
high confidencePurpose & Capability
The SKILL.md content matches the stated purpose: detailed guidance for nvidia-smi, NVML/pynvml usage, monitoring, troubleshooting, and basic process management. It references expected tools (nvidia-smi, watch, kill, pynvml). The registry metadata declared no required binaries — a minor documentation gap but not an incoherence: the referenced commands are appropriate to the stated purpose.
Instruction Scope
Instructions stay within the GPU monitoring/management scope. They advise running nvidia-smi, using watch/dmon, parsing CSV output, and include a Python pynvml monitoring example. They also recommend killing processes (kill <PID>) to free VRAM, which is a legitimate but potentially destructive admin action; the skill does not instruct reading unrelated files or exfiltrating data.
Install Mechanism
This is an instruction-only skill with no install spec or code files, which is the lowest-risk install model. The Python example references the external library 'pynvml' but no install instructions are provided — this is an omission for usability but not an active security risk from the skill bundle itself.
Credentials
The skill requests no environment variables, credentials, or config paths. It performs tasks that do not require secrets. This is proportionate to the stated purpose.
Persistence & Privilege
The skill is not marked always:true and does not request persistent system-level privileges. Note: the skill's runtime instructions include commands that terminate processes (kill) and run long-running monitoring; if the agent is allowed to execute shell commands autonomously, that capability could be used to perform destructive actions. Autonomous invocation itself is the platform default and is not a fault of the skill, but users should be aware.
Assessment
This is a coherent, instruction-only tutorial for nvidia-smi and NVML. Before installing or enabling it: 1) Confirm your system actually has NVIDIA drivers and nvidia-smi available; the skill assumes these tools exist. 2) Be aware the instructions include process termination (kill <PID>) — only allow the agent to run such commands if you trust it or review actions first. 3) The Python example requires the pynvml library; install it manually (pip install nvidia-ml-py3 or similar) if you plan to run the sample. 4) Because the skill can instruct shell actions, consider limiting autonomous execution or requiring user confirmation for destructive commands. No credentials are requested by the skill, which reduces risk.Like a lobster shell, security has layers — review code before you run it.
latestvk97006bczph0tzwv6ztashvvm9838wnz
License
MIT-0
Free to use, modify, and redistribute. No attribution required.