ml-audit-repo

What problem does it solve?

Evidence-driven ML repository verification prevents unverified claims from slipping into production by systematically auditing code, data, training, and metrics, and by producing verifiable artifacts.

Core Features & Use Cases

• Conducts staged audits (quick / standard / forensic) to verify executability, data reality, training authenticity, and metric validity.
• Generates verdicts, findings JSON, evidence manifest, and human-readable audit reports to support pre-merge and post-release checks.
• Enables teams to enforce reproducibility and traceable provenance for ML models and datasets.

Quick Start

Point the audit_skill at a repository path and choose an audit level to generate a complete evidence bundle and verdicts.