AR Policy Tester

Overview

Testing targets the two-step pipeline separately:

• Generated scenarios test rule correctness. They're derived from your rules and remove translation uncertainty. Review each: thumbs-up saves a SATISFIABLE test; thumbs-down → annotate.
• QnA tests test the full pipeline (translation + validation) with realistic question/answer pairs and an expected result.

Recommended order: scenarios first (fix rules), then QnA (fix translations / variable descriptions).

Reference: ../../shared/references/ar-api-context.md, findings-reference.md (severity ordering).

When to use

• After ar-policy-reviewer is clean, to validate behavior before deploying.
• To build a regression suite that re-runs whenever the policy changes.