If you're working on Kaggle competitions or want to learn from winning solutions, this extracts and organizes techniques from top performers across NLP, computer vision, time series, tabular, and multimodal domains. Feed it a competition URL and it pulls out detailed technical analysis from the top 20 solutions, including specific tricks, model configs, and code templates, all sorted into a structured knowledge base. The self-evolving aspect is clever: the more competitions you mine, the richer your reference library gets. It's essentially building you a personal playbook of what actually works in production ML competitions, which beats reading scattered forum posts when you're trying to move up the leaderboard.
npx -y skills add galaxy-dawn/claude-scholar --skill kaggle-learner --agent claude-codeInstalls into .claude/skills of the current project.
Select a file.
juliusbrussee/caveman
mattpocock/skills
shadcn/improve
obra/superpowers
forrestchang/andrej-karpathy-skills
vercel-labs/skills