ml-lab¶

A Claude Code plugin that runs structured ML hypothesis investigations using an adversarial critic-defender debate protocol or an independent ensemble of critics. It enforces rigor at every step — pre-specified metrics, confidence-tiered review findings, agreed experiments only — and produces a self-contained report with a production re-evaluation.

When to use ml-lab¶

Use ml-lab when you have a falsifiable ML hypothesis and want a structured process to test it. The plugin handles the full lifecycle: sharpening the hypothesis, building a minimal proof-of-concept, adversarial review (or ensemble sweep), pre-registered experiments, and synthesis of conclusions. It is designed for rigor over speed.

Quick start¶

# Install the plugin
/plugin marketplace add chris-santiago/ml-lab
/plugin install ml-lab@ml-lab

# Start an investigation
/ml-lab

Where to go next¶