enforcing-leakage-firewall

What problem does it solve?

This Skill helps you avoid trusting inflated test scores when a supervised model can leak information across entities, sources, or post-join features.

Core Features & Use Cases

• Flags random-split evaluations that hide duplicate rows, repeated customers, repeated patients, or other shared entities across splits.
• Recommends group-aware splitting, leave-one-source-out validation, and leave-one-feature-out checks to expose cross-source or target-derived leakage.
• Use it for multi-source crosswalks, recommender histories, merged registries, or any pipeline that shows suspiciously high performance after joins or feature engineering.

Quick Start

Use the enforcing-leakage-firewall skill to audit my model, identify leakage risks, and tell me whether the held-out metric is trustworthy.