skills.rest
raja21068
raja21068COMMUNITY
Explore all skills by this author →
WebsiteGitHub

monitor-experiment

Monitor running experiments and summarize JSON metrics with baseline deltas.

2|1|Updated Apr 19, 2026
Search Tags:#wandb#multi-agent workflow#baseline comparison#experiment monitoring#results aggregation#ml training logs#SSH screen sessions
One-click install
npx skills add https://github.com/raja21068/AutoResearch --skill monitor-experiment-raja21068
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: monitor-experiment
Source: https://github.com/raja21068/AutoResearch/tree/main/skills/aris/monitor-experiment
Command: npx skills add https://github.com/raja21068/AutoResearch --skill monitor-experiment-raja21068

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

It helps you determine whether long-running experiments have finished and whether they produced valid, comparable results instead of guessing from partial logs.

Core Features & Use Cases

  • Progress and output monitoring: Checks running screen sessions (or Modal runs) and captures recent log lines for quick status.
  • Result collection & validation: Locates JSON result files in the configured results directory, fetches the latest outputs, and summarizes key metrics.
  • Metrics enrichment (optional): Pulls Weights & Biases curves and run summaries when wandb: true is configured, improving insight into convergence and divergence.
  • Actionable interpretation: Compares metrics against the correct baseline, flags issues like negative deltas, NaNs, divergence, and recommends next steps.
  • Notification support (optional): Sends a Feishu notification when ~/.claude/feishu.json is configured and enabled.

Quick Start

Monitor the currently running experiments and summarize the latest JSON metrics with baseline deltas by running monitor-experiment with your server or screen name as the argument.

Frequently Asked Questions about monitor-experiment

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

FAQPage Schema
How do I check if my ML training experiments are still running via SSH screen sessions?▼

To check ML training experiment status via SSH screen sessions, inspect the running screen processes and capture recent log lines to determine completion progress and current execution state.

How do I aggregate JSON result files from multiple training runs for baseline comparison?▼

Aggregating JSON result files from multiple training runs involves locating the configured results directory, fetching the latest outputs, and calculating baseline deltas to summarize key metrics for comparison.

Can I pull Weights & Biases metrics for convergence and final evaluation analysis?▼

Yes, you can pull Weights & Biases metrics for convergence and final evaluation analysis by configuring the wandb option as true, which queries the W&B API to retrieve run curves and summaries.

What is the best way to detect divergence or NaN values in multi-agent workflow training logs?▼

The best way to detect divergence or NaN values in multi-agent workflow training logs is to aggregate outputs and apply guardrails that automatically flag negative deltas, NaNs, and divergence against the correct baseline.

Do I need a Feishu configuration file to send experiment monitoring notifications?▼

You need a Feishu configuration file located at ~/.claude/feishu.json to send experiment monitoring notifications, which must be configured and explicitly enabled to trigger alerts upon experiment completion.

Related Skills

Explore More Education & Research Skills→
Leo1349Leo1349

monitor-experiment

Monitor remote experiment sessions and summarize JSON result files.

Community
Intermediate
llv22llv22
5

analyze-results

Analyze ML experiment results to compute statistics and generate comparison tables.

Community
Intermediate
lightrain-alightrain-a

monitor-experiment

Monitor running ML experiments and retrieve logs, JSON results, and W&B metrics.

Community
Advanced
Shallow-WShallow-W

monitor-experiment

Monitor ML experiments across SSH, Vast.ai, and Modal deployments.

Community
Intermediate