skills.rest
OpenLAIR
OpenLAIROFFICIAL
Explore all skills by this author →
GitHub

aris-compute-guard

Verifies GPU and compute resource availability before running ML experiments.

1.1k|116|Updated Feb 26, 2026
Search Tags:#GPU#ssh#pre-flight#nvidia-smi#experiment validation#Machine learning#compute check
One-click install
npx skills add https://github.com/OpenLAIR/dr-claw --skill aris-compute-guard
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: aris-compute-guard
Source: https://github.com/OpenLAIR/dr-claw/tree/main/skills/aris-compute-guard
Command: npx skills add https://github.com/OpenLAIR/dr-claw --skill aris-compute-guard

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Running ML experiments without verifying compute resources wastes time and risks the AI fabricating fake results when execution silently fails. This Skill performs a mandatory pre-flight check that detects whether local GPUs, remote SSH servers, Vast.ai instances, or Modal serverless compute are actually available, and stops the pipeline immediately if they are not.

Core Features & Use Cases

  • Multi-Environment Detection: Reads the project's CLAUDE.md to determine the target environment (local CUDA, Apple Silicon MPS, remote SSH, Vast.ai, or Modal) and runs the appropriate availability check.
  • Hard Stop on Failure: When compute is unavailable, halts all experiment execution and reports exactly what is missing with actionable remediation steps, preventing hallucinated experiment results.
  • Resource Summary on Success: When compute is verified, prints GPU name, count, and free memory before proceeding.
  • Use Case: Before launching a training run on a remote GPU server, invoke this check to confirm SSH connectivity and free GPU memory, so the experiment never starts on an unreachable or fully occupied machine.

Quick Start

Ask the AI to run the compute guard check before starting your experiment, for example: "Check whether my GPU resources are available before running this training experiment."

Frequently Asked Questions about aris-compute-guard

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

FAQPage Schema
How do I check GPU availability before running ML experiments?▼

Run nvidia-smi to query GPU index, memory usage, and utilization, or use PyTorch to test torch.cuda.is_available() and MPS availability. A GPU counts as free when its used memory is below 500 MiB.

How to verify a remote GPU server is reachable over SSH?▼

Test connectivity with ssh -o ConnectTimeout=10 -o BatchMode=yes followed by a remote nvidia-smi query. If SSH fails due to network, authentication, or timeout, treat the server as unavailable and stop the experiment.

Does this compute check work on Apple Silicon Macs?▼

Yes, it detects Apple Silicon by checking torch.backends.mps.is_available() in PyTorch. If neither MPS nor CUDA is available, it warns that CPU-only execution is unsuitable for most ML training experiments.

Can I run experiments without a local GPU using Modal or Vast.ai?▼

Yes, set gpu: modal or gpu: vast in CLAUDE.md. Modal only requires an installed and authenticated CLI since it allocates GPUs serverlessly, while Vast.ai requires a running instance with SSH access.

What happens when no compute resources are available?▼

The check stops the entire experiment pipeline immediately and reports the specific issue, such as no GPU detected or SSH failure, along with remediation steps. It never fabricates or imagines experiment results.

Related Skills

Explore More Education & Research Skills→
lightrain-alightrain-a

run-experiment

Deploys ML experiments to local, SSH, Vast.ai, or Modal GPUs with verification.

Community
Advanced
gyy0592gyy0592
7

gen-report-concise

Generate concise experiment reports from training run artifacts and logs.

Community
Advanced
SciMate-AISciMate-AI
1

get-available-resources

Detect and summarize CPU, GPU, memory, and disk resources into a JSON report.

Official
Advanced
samuelmukotisamuelmukoti

modal-serverless-gpu

Run ML workloads on serverless GPU cloud with Python-native infrastructure definitions.

Community
Advanced