Verified:GitHub

/exp-run

Deploy and collect machine learning experiments from wiki plans across local and SSH environments.

77|16|Updated May 9, 2026

Search Tags:#machine learning #experiment execution #status monitoring #remote deployment #results collection #wiki automation

Ready to run

npx skills add https://github.com/Lambenthan/empiricalwiki --skill exp-run-lambenthan

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: /exp-run
Source: https://github.com/Lambenthan/empiricalwiki/tree/main/.claude/skills/exp-run
Command: npx skills add https://github.com/Lambenthan/empiricalwiki --skill exp-run-lambenthan

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

It solves the operational bottleneck of turning an experiment plan stored in wiki pages into actually running code, then reliably monitoring and collecting results back into the wiki.

Core Features & Use Cases

Three execution modes: default deploy (Phase 1–2), monitoring/collect (Phase 3), or fast full runs (Phase 1–4) for local experiments.
Deterministic experiment code generation & reproducibility scaffolding: writes a consistent code package under experiments/code/{slug}/ including train.py, config.yaml, run.sh, and optional requirements.txt, with checkpointing and multi-seed result saving.
Operational status management & reporting: updates wiki/experiments/{slug}.md status/outcome/results metadata and produces terminal reports for both deployment and collection, with support for local (GPU directly) and remote (SSH) environments.
Optional code review gate: runs an LLM code review during Phase 1 when --review is provided, improving correctness before deployment.

Quick Start

Tell the AI to deploy a planned experiment by running: /exp-run <experiment-slug> --env local.

Frequently Asked Questions about /exp-run

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

FAQPage Schema

How do I automate deploying and running machine learning experiments from a wiki?▼

To automate machine learning experiment deployment from a wiki, use a tool that reads experiment plans, generates reproducible code packages, and executes them on local or remote SSH environments. This process turns planned wiki pages into running code while monitoring status transitions automatically.

Can I run machine learning experiments remotely over SSH and collect results?▼

Yes, you can run machine learning experiments remotely over SSH and collect results. The execution process supports remote SSH deployments by generating code, starting runs, monitoring state, and collecting multi-seed result JSON files back into your wiki.

How do I collect and report experiment results with mean and standard deviation?▼

To collect and report experiment results with mean and standard deviation, execute a collection phase that gathers per-seed JSON files from the results directory. It aggregates metrics and summarizes them into the wiki with per-metric mean±std reporting.

What is the best way to ensure reproducibility when generating machine learning experiment code?▼

The best way to ensure reproducibility when generating machine learning experiment code is to scaffold a consistent package including train.py, config.yaml, and run.sh. This deterministic generation supports checkpointing and multi-seed result saving for reliable reproduction.

Does the experiment execution process support code review before deployment?▼

Yes, the experiment execution process supports an optional code review gate before deployment. By enabling a review flag during the code generation phase, an LLM reviews the generated code to improve correctness before the run starts.

Can I deploy only the experiment code without collecting results immediately?▼

Yes, you can deploy only the experiment code without collecting results immediately. The execution process supports deploy-only, collect-only, and full execution modes, allowing you to separate code deployment from result collection.