run-experiment

Deploy and run ML experiments on local or remote GPU servers.

Updated Apr 10, 2026
One-click install
npx skills add https://github.com/KwongFuk/codex-skills --skill run-experiment-kwongfuk
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: run-experiment
Source: https://github.com/KwongFuk/codex-skills/tree/main/global/run-experiment
Command: npx skills add https://github.com/KwongFuk/codex-skills --skill run-experiment-kwongfuk

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Deploy and run ML experiments on local or remote GPU servers, enabling researchers and engineers to kick off training jobs from a single command set.

Core Features & Use Cases

  • Environment-aware deployment: reads AGENTS.md to determine remote versus local context and available GPUs.
  • Pre-flight GPU checks ensure GPUs are available before launching training tasks.
  • Code synchronization supports rsync or git workflows to push code to the target machine.
  • Optional WandB integration can be auto-instrumented in training scripts when configured.
  • Remote deployment uses SSH with a screen session to run training, while local deployment runs directly on the host.
  • Verification and optional Feishu notifications can report status.

Quick Start

Begin by providing your experiment script and an AGENTS.md configuration to deploy and run the training on the target GPU server or locally.

Frequently Asked Questions about run-experiment

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

FAQPage Schema
How do I run ML experiments on a remote GPU server via SSH?

The tool orchestrates remote ML experiment deployment by using SSH to connect and run training tasks within a screen session. It reads AGENTS.md for configuration and performs pre-flight GPU checks before launching the job.

What is the best way to synchronize code to a remote GPU for training?

The best way to synchronize code for remote GPU training is using either rsync or git workflows. The tool handles this code transfer automatically as part of its deployment orchestration before launching the experiment.

Does this tool support automatic WandB integration for experiment tracking?

Yes, the tool supports WandB integration by automatically instrumenting your training scripts when configured. This allows you to track ML experiments without modifying your code.

Can I use this to check GPU availability before launching training scripts?

Yes, you can use this to check GPU availability because it performs pre-flight GPU checks before launching training tasks. This ensures GPUs are available on the target machine before deployment.

How do I configure local versus remote deployment for my ML experiment?

You configure local versus remote deployment by providing an AGENTS.md file. The tool reads this configuration to determine whether to run directly on the host or deploy across a remote SSH session.

Why do I need an AGENTS.md file to launch my training jobs?

You need an AGENTS.md file because it provides the environment-aware configuration required to determine remote versus local context and available GPUs. It acts as the central setup file for orchestrating the deployment.