AutoCedar is a verifier-guided agentic framework for synthesizing Cedar authorization policies from natural-language requirements. It first decomposes schema and policy authoring into small, reviewable intent atoms describing vocabulary and behavior. After mechanical validation and human review, a language model proposes a policy, while a verifier checks it against the approved target. Failures become repair signals indicating whether the policy should be broadened, narrowed, or restructured without changing the target. The paper reports convergence on all 221 CedarBench tasks and presents case studies in healthcare, education, and conference management.
No heat snapshots are available in the last 24 hours.