Experiment Pipeline

This is a structured workflow for running machine learning experiments in four mandatory stages: reproduce the baseline, tune hyperparameters, validate your novel method, and run ablations. Each stage has an attempt budget (20, 12, 12, 18) and a gate condition you must pass before moving forward. The discipline is the point here. It forces you to verify your baseline actually works before you start testing your fancy new architecture, and it stops you from burning 50 GPU hours on hyperparameter searches that should have taken 12 runs. Integrates with other EvoScientist skills for loading prior experimental strategies and diagnosing failures. If you tend to jump straight to testing your idea without validating the baseline setup first, this will save you from yourself.

skills.sh installs: 396
GitHub stars: 488
last updated: Aug 21, 2026

View on GitHub

Install

$ npx -y skills add evoscientist/evoskills --skill experiment-pipeline --agent claude-code

Installs into .claude/skills of the current project.

Files

Experiment Pipeline

A structured 4-stage framework for executing research experiments from initial implementation through ablation study, with attempt budgets and gate conditions that prevent wasted effort. This follows the Experiment Tree Search design from the EvoScientist paper, where the engineer agent iteratively generates executable code, runs experiments, and records structured execution results at each stage.

When to Use This Skill

User has a planned experiment and needs to organize the execution workflow
User wants to systematically validate a novel method against baselines
User asks about experiment stages, attempt budgets, or when to move on
User needs to reproduce baseline results before testing their method
User mentions "experiment pipeline", "baseline first", "ablation study", "stage budget", "experiment execution"

The Pipeline Mindset

Experiments fail for two reasons: wrong order and no stopping criteria. Most researchers jump straight to testing their novel method without verifying their baseline setup, then wonder why results don't make sense. Others spend weeks tuning hyperparameters without a budget, hoping the next run will work.

SKILL.md on GitHub