Claude Skills by msilverblatt

Use after every experiment run. This is the core data science skill — reading results, understanding errors, and forming the next hypothesis. If you skip this, you're not doing science, you're doing random search. ---

Use when generating feature hypotheses from domain knowledge. This is not a one-time pre-work step — return here whenever results surprise you, progress stalls, or a new data source becomes available. ---

Use when you need to build intuition about the data. This isn't a checkbox — it's how you generate your first hypotheses and catch problems before they poison your models. ---

Use before creating any experiment. This is the thinking step. If you skip it, you'll run experiments that don't teach you anything. ---

Use when creating and testing new features. Every feature is a hypothesis about the data. Treat it that way. ---

Load this skill once at the start of any ML session. It sets the frame for everything else. ---

Use when working on ML experimentation tasks — data preparation, feature engineering, model selection, hyperparameter tuning in an harnessml project. Guides the complete ML workflow with sound data science practices.

Use when evaluating your model ensemble — what's in it, what's missing, and whether the models are actually providing diverse perspectives on the problem. ---

Use when starting a new ML project or revisiting the scope of an existing one. ---

Use when executing an experiment. Load `experiment-design` first to ensure you've thought through the hypothesis. This skill covers the mechanics and the discipline of execution. ---

Use after running several experiments, when you need to step back and connect the learnings. This is also the skill for deciding what to do next — and for recognizing when further experiments aren't adding understanding. ---