Skip to content
Get packs

Monitor Experiment Results

Check on running ML experiments on remote servers and summarize their results.

Source Download skill
Repository
wanshuiyin/Auto-claude-code-research-in-sleep
Stars
18,482
Licence
MIT
Files
1 file

What it does

When you ask an agent to check on experiments, this skill tells it how to look at jobs running on remote machines (SSH servers, Vast.ai GPU rentals, or Modal), read their screen sessions and log output, pick up JSON result files and Weights & Biases training metrics, and then present the numbers in a comparison table with a short interpretation. It also reminds you to shut down paid Vast.ai instances when work is finished, and can send a notification if you have configured one.

Great for

  • ML researchers who run training jobs on remote GPU servers and want a quick status report
  • People who use Vast.ai or Modal and want reminders about cost and idle billing
  • Teams that already log experiments to Weights & Biases and want the agent to pull curves and summaries

Skip it if

  • Users without remote servers or experiment runs, since the workflow assumes SSH access and existing jobs (not useful outside that setup)
  • Judging whether research results are scientifically good: the skill only flags obvious problems like NaN, divergence or negative deltas, and it is meant for machine-checkable status rather than quality
  • Situations where the server or credentials are shared or sensitive, because the agent will run remote commands over SSH and read job outputs

Try asking

  • “Check whether my training run on the lab server is finished and show the latest results”
  • “Is the vast.ai experiment done? Summarize the metrics against the baseline”
  • “Monitor the Modal job and tell me the execution time and estimated cost”
  • “Pull the latest W&B loss curves for my run and tell me if it is diverging”
Step-by-step guide: how to use Monitor Experiment ResultsInstall, first run, examples, tips and FAQs

Install in 3 steps

  1. 1

    Download the skill

    The ZIP contains the 1 original files, the MIT licence and an attribution note.

    monitor-experiment.zip
  2. 2

    Add it to your agent

    Claude Code (all projects)

    mkdir -p ~/.claude/skills && unzip monitor-experiment.zip -d ~/.claude/skills/

    Claude Code / Agent SDK (this project only)

    mkdir -p .claude/skills && unzip monitor-experiment.zip -d .claude/skills/

    Claude apps (web/desktop): Settings → Capabilities → Skills → Upload skill, and choose the ZIP.

  3. 3

    Ask for the task

    No special command needed — the agent reads the skill's description and loads it when your request matches. Try one of the prompts above.

Get better results

  • Give the server alias or screen session name as the argument so the agent knows which job to check
  • Keep your CLAUDE.md up to date with wandb and gpu settings, since the skill branches on them
  • Review commands the agent proposes over SSH before approving, since it runs remote shell and Python snippets on your servers

What's inside

  • SKILL.md

Redistributed unmodified from wanshuiyin/Auto-claude-code-research-in-sleep under the MIT licence. The assessment was drafted with AI from the skill's files on 2026-10-29. Skills can instruct your agent to run code — read SKILL.md and any scripts before installing.