6 skills.
Use when asked to define or document data governance — the policies, ownership, and processes for how an organization manages its data as an asset (who owns a dataset, how it's classified, how long it's retained, who can request access) — as distinct from [[data-quality]], which is the measurable property those policies aim to protect, not the framework itself.
Use when asked to design, document, or troubleshoot a data pipeline — an automated sequence that extracts data from a source, transforms it, and loads it into a destination on a schedule or trigger — as distinct from [[data-quality]], which measures whether the data moving through the pipeline is actually accurate and complete, not how it's moved.
Use when asked to define, measure, or improve data quality — whether data is accurate, complete, consistent, and fit for its actual use — as distinct from [[data-governance]], which is the policy and ownership framework around data, not the measurable property of the data itself.
Use when asked to write or review a model card — a document accompanying a trained ML model that describes what it does, how it was trained and evaluated, and its known limitations — as distinct from [[mlops]], which covers the infrastructure for deploying and monitoring a model rather than documenting it for downstream users.
Use when asked about MLOps — the practices and infrastructure for deploying, monitoring, and maintaining ML models in production reliably — as distinct from the model-development and training work itself, and from [[machine-learning-model-card]], which documents a model for downstream users rather than operating it once it's live.
Use when asked to write, improve, or debug a prompt for a language model — getting reliable, intended behavior through specific instructions, examples, and context — as distinct from [[machine-learning-model-card]], which documents a trained model's capabilities and limits rather than shaping its behavior through the prompt.