ExpAgent

This skill should be used when the user asks to "create a new experiment", "start an experiment", "新しい実験", "実験を作って", "run training", "train a model", "record results", "結果を記録", "plan next experiment", "次の実験を考えて", "review experiment history", "実験の履歴を見て", or wants to follow the experiment lifecycle (plan, create, implement, train, record).

This skill should be used when the user asks to "create a task", "list tasks", "show the board", "check project status", "add a milestone", "create a document", "record a decision", "search backlog", "initialize backlog", "show overview", "manage drafts", "update task status", or any project management operation using the backlog CLI. Also use this skill when the user starts working on a task, finishes a task, or wants to check what to work on next — backlog is the single source of truth for project context across sessions. Provides workflows for the Backlog.md project management system.

This skill should be used when the user asks to "setup competition", "setup infra", "GCP setup", "GCPセットアップ", "初期セットアップ", "コンペのセットアップ", "インフラ構築", "データダウンロード", "download competition data", "setup environment", "環境構築", or wants to initialize infrastructure and download competition data for a new competition project.

Comprehensive primary skill for agents working with Weights & Biases. Covers both the W&B SDK (training runs, metrics, artifacts, sweeps, reports) and the Weave SDK (GenAI traces, evaluations, scorers). Includes helper libraries, gotcha tables, and data analysis patterns. Use this skill whenever the user asks about W&B runs, Weave traces, evaluations, training metrics, loss curves, model comparisons, or any Weights & Biases data — even if they don't say "W&B" explicitly.

This skill should be used when the user asks to "submit to kaggle", "kaggle submit", "push submission", "submit experiment", "サブミット", "sub して", "提出して", "kaggle に提出", "check submission results", "結果確認", "スコア確認", "public score", "LBスコア記録", "submission failed", "サブミット失敗", or wants to run the Kaggle submission pipeline (deps push, model upload, kernel push, result check, troubleshooting).