What problem does it solve? Machine learning models often pass paper metrics yet fail in production due to data drift, miscalibration, or undocumented assumptions. This Skill provides an independent, evidence-based audit of models across their full lifecycle, producing severity-rated findings and audit-grade reports. ## Core Features & Use Cases - End-to-End Model Audit: Covers 10 domains from documentation and governance review through data reconstruction, feature analysis, replication, calibration, performance, fairness, and business impact. - Statistical Testing Toolkit: Includes ready-to-use Python implementations for PSI, Gini/KS discrimination metrics, Hosmer-Lemeshow calibration tests, SHAP global/local analysis, and Partial Dependence Plots. - Audit-Grade Reporting: Generates a structured QA report with severity-rated findings (High/Medium/Low/Info), quantified business impact, and remediation tracking. - Use Case: A bank's risk team needs to validate a credit scoring model before regulatory review. Use this Skill to reconstruct the modeling population, replicate training, test calibration across segments, run SHAP fairness analysis, and deliver a governance-ready QA report. ## Quick Start Audit my XGBoost credit risk model by replicating the training pipeline, testing calibration with Hosmer-Lemeshow, and producing a severity-rated QA report.
