Deep Agents adds rubric-grader integration hooks
LangChain released Deep Agents 0.7.11 with software-development-kit integration hooks for rubric graders. The change gives agent harnesses a defined point to attach structured evaluation logic rather than treating grading as an unrelated post-processing step.
Why it matters: Evaluation quality depends on where graders run, which evidence they see, and how failures propagate. A harness-level hook makes those choices versionable and testable alongside agent behavior.