experiment-monitor

What problem does it solve?

This Skill automates monitoring of active experiments by reading logs in logs/experiments/*.yaml, verifying process health, and surfacing up-to-date status to reduce manual chasing of stalled runs.

Core Features & Use Cases

• Load active experiments from logs/experiments/*.yaml and monitor their status.
• Check process liveness locally or remotely, tail recent logs, and extract key metrics (loss, accuracy, epoch, step).
• Detect errors such as OOM, NaN, or tracebacks; update status to running/completed/failed; optionally notify.

Quick Start

Monitor all active experiments and report their latest status in a single glance.