What happens if I run an experiment outside of a Git repository?

The skill will still log your parameters and results, but it will provide a warning that reproducibility is reduced because the code state cannot be pinned to a specific commit.

How does Experiment Log ensure reproducibility?

It snapshots the exact execution context, including the Git commit hash, dirty status, environment variables, and specific package versions used during the run.

Where are my experiment results stored?

Results are stored in two places: a local JSON log file for persistence and a knowledge graph for quick searching and cross-experiment comparisons.

Can I compare current runs with previous experiments?

Yes, the skill automatically queries your history to generate comparison tables that highlight how parameter changes have influenced your primary metrics.

Experiment Log

Name: Experiment Log
Author: jaminitachi

byjaminitachi

0•

データサイエンスとML

Tracks experiment parameters, results, and environment snapshots to ensure full reproducibility and systematic iteration.

Experiment Log is a specialized Claude Code skill designed to eliminate lost insights by systematically documenting every trial, parameter sweep, and result. it automatically captures the execution context—including Git commit hashes, environment variables, and package versions—to create a persistent, queryable history. By comparing current runs against previous data stored in a knowledge graph, it helps developers and data scientists identify trends, avoid repeated failures, and maintain a rigorous reproducibility trail for all technical experiments and model evaluations.

主な機能

010 GitHub stars

02Linkage between experiments, hypotheses, and related academic papers

03Automatic generation of unique experiment IDs and timestamped logs

04Persistent storage in local JSON logs and searchable knowledge graph entities

05Automated comparison tables across historical runs to identify regressions or improvements

06Environment snapshots including Git commit hash, branch, and package versions

ユースケース

01Maintaining a searchable database of technical trials to prevent redundant failures

02Tracking model performance metrics across different hyperparameter configurations

03Documenting the impact of specific code optimizations on system benchmarks

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add jaminitachi/superclaw experiment-log

For use in Claude.ai and ChatGPT

Download Skill