agentic-kaggle-skill
π€ AI Agent-driven Kaggle competition workflow. Battle-tested patterns for score stabilization, submission troubleshooting, kernel workflows, and spec-driven de
What is agentic-kaggle-skill?
π€ AI Agent-driven Kaggle competition workflow. Battle-tested patterns for score stabilization, submission troubleshooting, kernel workflows, and spec-driven development.
Overview of agentic-kaggle-skill
What it is
Agentic Kaggle Skill is a reusable workflow for AI agents to participate in Kaggle competitions. It provides battle-tested patterns for score stabilization, submission troubleshooting, kernel workflows, and spec-driven development. The skill is written in an open agent skills format and can be used by multiple agents, including Codex and Hermes.
Key features
- Reads Kaggle rules, data terms, metric, submission format, and scoring mode before modeling.
- Builds metric-correct baselines with stable folds and out-of-fold predictions.
- Uses public notebook and discussion intelligence as scouting signals, not copied source.
- Offloads heavy work to Kaggle notebooks/scripts when local compute is insufficient.
- Supports staged producer/consumer notebook pipelines with private artifact datasets.
- Handles code-competition hidden rerun failures, vague scoring errors, timeouts, and OOMs.
- Tracks reproducibility artifacts, run logs, score receipts, and ensemble evidence.
Use cases
Transforming Kaggle work from scattered manual iteration into an agent-assisted competition loop. It can be used for validation, modeling, Kaggle execution, debugging, ensembling, and scored submission. The skill is distilled from real Kaggle competition experience, including RL Game AI, Audio Classification, and LLM Reasoning.
Getting started
Install as a user-level Codex skill:bashmkdir -p ~/.agents/skillsgit clone https://github.com/FrankS-IntelLab/agentic-kaggle-skill.git ~/.agents/skills/agentic-kaggle-skillRestart Codex if the skill does not appear immediately. Invoke it explicitly with:textUse $agentic-kaggle-skill to help me start this Kaggle competition.For Hermes, install the whole skill folder:bashmkdir -p ~/.hermes/skills/data-sciencegit clone https://github.com/FrankS-IntelLab/agentic-kaggle-skill.git ~/.hermes/skills/data-science/agentic-kaggleMost scripts use only the Python standard library. scripts/make_folds.py requires pandas and numpy; scikit-learn is recommended for standard splitters.bashpython3 -m pip install -r requirements.txt
License and pricing
This project is licensed under the MIT License. It is open source and free to use.
Based on the project's public README and repository metadata.