What problem does it solve? Vague goals like "we need better quality data" leave producers and consumers of research datasets without shared, enforceable expectations. This Skill replaces undefined expectations with named quality dimensions, numeric objectives, and executable checks, producing a negotiated SLO document per dataset that pipelines can enforce. ## Core Features & Use Cases - SLI/SLO/SLA framework: Walks the seven EDM Council data-quality dimensions (accuracy, completeness, conformity, consistency, coverage, timeliness, uniqueness) crossed with service-level indicators like availability, freshness, retention, and incident clocks (Td/Tn/Tr). - Enforcement mapping: Maps every SLO to executable checks using SodaCL or Great Expectations, wired into Airflow/Argo orchestration, with per-check violation policies (halt, quarantine, or segment). - Negotiated SLO document: Produces a structured template covering parties, objectives, intrinsic/extrinsic splits, enforcement tables, lifecycle policies, and review cadence. - Use Case: A research team needs to guarantee that an evaluation set stays frozen and free of train/eval contamination. Use this Skill to define a zero-overlap uniqueness SLO with a pre-publish contamination scan that halts publication on violation. ## Quick Start Use the dataset-qos-slo-designer skill to draft an SLO document with executable quality checks for our training corpus covering label accuracy, deduplication, and freshness.
