AgentTrustCI: Build-time policy, grounding and trajectory testing for Java LLM agents

AgentTrustCI is a Java/JUnit 5 framework that validates LLM agents' tool selection, answer grounding and process adherence at build time using policy, grounding and trajectory checks. On 55 synthetic cases across seven enterprise domains, the full pipeline correctly classified all 30 injected faults and 25 passing cases, with none of the 12 key cases missed. The authors state the results only validate regression case coverage, not production accuracy.