:probabl.
Official@probabl-ai · France
The Tabular AI company by the creators of scikit-learn
Agent Skills by :probabl.
Showing 14 vetted skills indexed across 1 GitHub repositories.
probabl-aitest-ml-pipeline
Routes ML experiment test creation to category subskills while enforcing stem-pairing rules.
probabl-aiorganize-ml-workspace
Scaffolds and organizes ML experimentation project layouts with src packages, experiment scripts, and journals.
probabl-aiexplore-ml-data
Executes read-only exploratory data analysis and writes persisted EDA reports before model design.
probabl-aipython-api
Look up installed Python package APIs via inspect, pydoc, LSP, and versioned docs, then cache findings.
probabl-aiaudit-ml-pipeline
Executes read-only jupytext audit files against skore reports and renders markdown digests.
probabl-aipython-code-style
Runs ruff lint and format checks on edited Python files and enforces numpydoc docstrings.
probabl-aismoke-test-ml-pipeline
Writes diagnostic pytest smoke tests that verify ML pipeline predict-time row-count correctness.
probabl-aidata-science-python-stack
Selects and installs Python libraries for data science and ML projects by tier.
probabl-aievaluate-ml-pipeline
Evaluates sklearn-compatible learners with skore reports and data-driven cross-validator selection.
probabl-aibuild-ml-pipeline
Declare machine learning pipelines as skrub DataOps graphs from data source to predictor.
probabl-aipython-env-manager
Detects the project's Python environment manager and installs packages with the correct manager-specific commands.
probabl-aiiterate-ml-experiment
Orchestrates the propose-approve-implement-record loop for ML experiments via journal design notes.
probabl-aiiterate-from-user
Synthesizes user-provided articles, issues, or ideas into confirmed ML experiment proposals.
probabl-aiiterate-from-skore
Converts skore audit digest findings into ML experiment backlog candidates.
Frequently Asked Questions About :probabl.
FAQPage SchemaWhat tasks can I accomplish with probabl-ai's skills?▼
You can scaffold ML workspaces, run EDA with skrub TableReport, declare pipelines as skrub DataOps graphs, evaluate learners via skore cross-validation reports, write structural smoke tests, audit finished experiments, and iterate experiments through journal design notes and backlog sourcing.
Who are these skills designed for?▼
Data scientists and ML engineers working on tabular Python projects with the scikit-learn ecosystem. They suit practitioners who want disciplined experiment tracking, paired tests, and reproducible evaluation using skrub, skore, pytest, and ruff rather than ad-hoc notebooks.
How do the skills run in practice?▼
Skills trigger contextually: organize-ml-workspace scaffolds folders, iterate-ml-experiment drives the propose-approve-implement loop, and audit files execute via a bundled in-process runner (run_cells.py) using IPython. Python execution goes through scratch files with pixi, never inline python -c.
What are the prerequisites and dependencies?▼
A Python data-science environment managed by pixi (default), uv, poetry, or conda, detected by python-env-manager. Core libraries include scikit-learn, skrub, skore, pandas or polars, pytest, ruff, and jupytext; the agent feature additionally requires ipython and pyright.
How do the skills enforce experiment quality?▼
Design notes must be approved before experiment scripts exist, smoke tests assert prediction row counts match the predict grid to prove X-marker placement, ruff enforces lint and numpydoc style, and skore audit digests convert diagnostic issues into cited backlog rows.