Bedrock ships six skills for formal AI policies

Amazon Bedrock released 6 open-format Agent Skills covering the Automated Reasoning policy lifecycle: build, review, test, debug, deploy, and runtime validation. AWS’s worked example extracted six rules, four variables, and one custom type from three policy sentences, then attached a versioned policy to a guardrail; AWS cautions that rule extraction is nondeterministic and that solver certainty depends on faithful translation into formal logic. Teams adopting policy enforcement inside the agent harness should treat these skills as reviewable policy tooling, not proof that an agent-generated policy captured the source correctly.