skills.rest
Kirrito-k423
Kirrito-k423COMMUNITY
Explore all skills by this author →
WebsiteGitHub

server-hardware-probe

Probe remote GPU and NPU servers over SSH and parse npu-smi or nvidia-smi outputs.

Updated Jun 6, 2026
Search Tags:#gpu#ssh#utilization#paramiko#nvidia-smi#npu-smi#hardware probe
One-click install
npx skills add https://github.com/Kirrito-k423/AutoResearch --skill server-hardware-probe-kirrito-k423
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: server-hardware-probe
Source: https://github.com/Kirrito-k423/AutoResearch/tree/main/.agents/skills/03-server-hardware
Command: npx skills add https://github.com/Kirrito-k423/AutoResearch --skill server-hardware-probe-kirrito-k423

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires paramiko, and includes scripts (resource) components.

What problem does it solve? Collecting accelerator inventory and utilization from remote training servers requires manual SSH sessions and ad-hoc parsing of npu-smi or nvidia-smi output, which is slow and error-prone when checking many machines. ## Core Features & Use Cases - Remote Hardware Probing: Connects to remote GPU/NPU servers over SSH using paramiko and runs hardware query commands. - Output Parsing: Parses npu-smi info and nvidia-smi output to extract accelerator inventory, driver versions, memory and core utilization. - Process Occupancy Listing: Lists processes occupying accelerators via nvidia-smi --query-compute-apps=pid,process_name,used_memory. - Use Case: Before launching a training job, probe the target server to confirm free NPU memory and identify which processes currently occupy the devices. ## Quick Start Ask the agent to probe the hardware of a remote server such as nvidia-01 using the hardware_probe.py script with your config file and report accelerator inventory and occupancy.

Frequently Asked Questions about server-hardware-probe

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I check GPU or NPU usage on a remote server over SSH?▼

Run the hardware_probe.py script with your config file and target server name. It connects via SSH using paramiko, executes npu-smi or nvidia-smi remotely, and parses the output into structured hardware data.

How to list processes occupying a GPU with nvidia-smi?▼

Use nvidia-smi with --query-compute-apps=pid,process_name,used_memory to list compute processes and their memory usage. This skill runs that query remotely over SSH and parses the occupancy results.

Does this skill support both Ascend NPU and NVIDIA GPU servers?▼

Yes. It parses npu-smi info output for Ascend NPU servers and nvidia-smi output for NVIDIA GPU servers, covering accelerator inventory, driver versions, and utilization metrics for both platforms.

What information does npu-smi parsing extract?▼

The parser extracts accelerator inventory, driver versions, memory and core utilization, and process occupancy from npu-smi output. It also handles partial hardware data when some fields are unavailable.

When should I not use the server hardware probe?▼

Do not use it for network speed testing, service reachability checks, training stack health validation, or report rendering. Those tasks belong to separate skills in the workflow; this skill only covers SSH-based hardware probing and smi output parsing.

Related Skills

Explore More Software Engineering Skills→
ibleaiblea
4

ssh

Execute bash commands on remote servers via SSH connections.

Community
Intermediate
devrelaicomdevrelaicom
37

proof-server:proof-server-operations

Inspect readiness and job queue metrics for Midnight proof servers.

Official
Advanced
framersaiframersai

healthcheck

Probe endpoints to validate uptime, latency, DNS, and TLS validity.

Official
Advanced
mishahaninmishahanin
10

bridge-health

Aggregate liveness probes and telemetry to assess bridge daemon fleet health.

Community
Advanced