ml-docs

name	ml-docs
description	On-demand ML/data science library expert. Use when the user asks how to use any function, class, or method from NumPy, Pandas, scikit-learn, Matplotlib, TensorFlow, Keras, PyTorch, Seaborn, SciPy, statsmodels, XGBoost, LightGBM, Hugging Face Transformers, OpenCV, NLTK, spaCy, Plotly, Dask, PySpark, SQLAlchemy, or Jupyter. Fetches and synthesizes official API docs, parameter reference, and working code examples. Also use when the user asks "how do I do X in pandas/numpy/torch/sklearn", needs to understand a deep learning layer or training loop, asks about NLP pipelines, computer vision transforms, statistical tests, SQL ORM patterns, or big data ops.
allowed-tools	Bash(uv run * scripts/process.py *) Read Write WebFetch Glob Grep
argument-hint	[library] [topic or function name]
model	haiku
effort	low
compatibility	>=1.0
metadata	{"category":"reference","tags":["docs","api-reference","numpy","pandas","pytorch","sklearn","huggingface","xgboost","lightgbm","statsmodels"],"phase":"any"}

ML Docs

Context: $ARGUMENTS

Quick start

• Look up a specific function/class: → Step 1, then Step 2
• Understand a concept or workflow: → Step 1, pick a topic URL, fetch it
• Compare options across libraries: → run Step 1 for each library, fetch in parallel

When to use

• User asks how to use any function, class, or method from a supported library
• User asks "what parameters does X take" or "what does Y return"
• User needs working code examples for a task (groupby, cross-validation, fine-tuning, etc.)
• User asks about a concept: broadcasting, autograd, attention, sparse matrices, etc.
• User wants to know which library/function to use for a task
• User hits an error and needs to check expected behavior from the official docs

Step 1 — Resolve the documentation URL

Run process.py resolve to get the prioritized list of URLs to fetch:

uv run ${CLAUDE_SKILL_DIR}/scripts/process.py resolve \
  --library <library-name-or-alias> \
  --query "<function, class, or topic>"

The script returns a JSON object with fetch_in_order — a list of URLs ranked by specificity. Start with priority 1. If it returns 404 or empty content, move to priority 2, then 3.

Library aliases accepted: See references/guide.md for the full alias table.

Examples:

uv run ${CLAUDE_SKILL_DIR}/scripts/process.py resolve --library pandas --query "DataFrame.groupby"
uv run ${CLAUDE_SKILL_DIR}/scripts/process.py resolve --library torch --query "autograd"
uv run ${CLAUDE_SKILL_DIR}/scripts/process.py resolve --library sklearn --query "cross_val_score"
uv run ${CLAUDE_SKILL_DIR}/scripts/process.py resolve --library transformers --query "pipeline"
uv run ${CLAUDE_SKILL_DIR}/scripts/process.py resolve --library scipy --query "hypothesis testing"

To list all supported libraries:

uv run ${CLAUDE_SKILL_DIR}/scripts/process.py list