●builderYou can use this harness to implement more reliable, verifiable agentic reasoning through executable hypotheses.
●researcherThis provides a method for evaluating agents in interactive environments using formal transition checks rather than just pattern matching.