diyu-eval-harness

AI behavior evaluation is automated by running a harness that identifies and quantifies governance, capability, and regression metrics across registered objects, producing scorecards and traceable decision artifacts.

A governance-focused AI capability evaluation verifies the presence and integrity of governance artifacts and ensures alignment with upstream references. You need it when assessing declared capabilities against actual behavior for objective auditing.

To generate a 10-point scorecard, invoke the evaluation harness to run checks across all registered objects, computing a six-dimension scorecard that emits both Markdown reports and YAML artifacts.

Yes, the evaluation harness exports machine-readable YAML scorecards alongside human-readable Markdown reports, providing evidence and auditing artifacts suitable for governance reviews and archival compliance.

Yes, you can detect degradation or improvement by running a regression evaluation that compares current AI behavior results against an established baseline across your registered objects.

The best way to evaluate eight registered AI objects is to invoke the end-to-end evaluation harness, which executes governance, capability, and regression workflows and generates a six-dimension, 10-point scorecard for review.