from agentic-kaggle-skill53
Assistant skill to run and manage Kaggle competition workflows: submission troubleshooting, score stabilization patterns, kernel workflows, and spec-driven dele
This skill packages practical, battle-tested patterns for running Kaggle competitions with AI agents. It teaches the agent how to set up projects, automate routine monitoring, prepare and push kernels, handle common submission errors, and use spec-driven development to delegate code and experiments to coding agents. The goal is reliable, repeatable competition workflows that reduce manual overhead and improve leaderboard performance.
Use this skill when starting or maintaining a Kaggle competition project, when submissions fail or return unexpected results (400 errors, format issues), when you need to reproduce top notebooks, or when you want to safely delegate GPU training and environment setup to subagents. It’s also useful for stabilizing noisy leaderboard scores by following time-based evaluation guidance.
kaggle CLI commands and kernel management.Best used with coding/delegation-capable agents (OpenCode, Claude Code, or subagent delegates) for tasks like pushing kernels, editing SPEC.md, and orchestrating training runs.
This skill has not been reviewed by our automated audit pipeline yet.
