audit-ml-pipeline

To generate an audit report, load a Skore project report using bare Python expressions to produce a detailed markdown digest. The output summarizes checks and metrics with actionable mitigation URLs for further inspection.

A post-experiment audit reviews machine learning metrics and checks to ensure model validity. This Skill loads Skore project reports to create a human-readable markdown digest, establishing an audit trail for backlog creation.

Yes, running a Skore project audit requires proper versioning of skore libraries and read-only access to the skore Project. You also need access to the report documentation URL to generate the markdown digest.

Yes, the audit process supports re-execution for reviewing previous experiment audits. Re-running loads the Skore report again to output a summary of checks and metrics with actionable mitigation URLs.

Automating ML pipeline audits requires skore, skrub, sklearn, and sklearn-utils. These dependencies enable loading Skore reports and generating human-readable markdown digests for analysis.

The best way to create an audit trail is loading Skore project reports to generate a markdown digest. This creates a readable trail that the skore source branch can read for backlog creation.