/exp-run

What problem does it solve?

This Skill runs end-to-end machine learning experiments defined in your wiki, turning a planned experiment plan into deployed code, tracked execution, and collected results—while enforcing a manual approval gate before any deployment or execution.

Core Features & Use Cases

• Experiment code generation & deployment: Converts wiki/experiments/{slug}.md into runnable code under experiments/code/{slug}/, then deploys it locally or remotely.
• Operational safety via manual inspection: Requires user confirmation and asks for manual checks of critical configuration (dataset paths, interfaces, API settings) before execution proceeds.
• Three execution modes: Supports deploy-only (default), collect-only (--collect / --check), and full end-to-end (--full) workflows.
• Monitoring & result collection: Collects metrics and summarizes outcomes into wiki/experiments/{slug}.md, while appending operational logs to wiki/log.md.

Quick Start

Run: /exp-run <experiment-slug> --env local