Your pipeline deploys multi-agent code to Azure but only runs unit tests on pure Python helpers. Several regressions shipped because tool wiring and prompt changes were untested. What CI/CD requirement should you add?
Select an answer to reveal the explanation.
Short Explanation
You need automated integration/regression tests and agent evaluations for critical tool and prompt scenarios as release gates, alongside unit tests and infrastructure-as-code validation. Python-only unit tests won't catch broken tool graphs or prompt behavior. Removing tests, skipping infrastructure checks, or allowing force-pushes without review just increases risk.
Full Explanation
Specify CI/CD requirements for integration and release: infrastructure-as-code validation, testing layers, and automated evaluations that exercise multi-agent paths. Use automated integration/regression tests and agent evaluations, including critical tool and prompt scenarios, as release gates alongside unit tests and infrastructure validation. Fail the pipeline when groundedness, tool choice, or safety scores regress. Removing unit tests to speed the pipeline is wrong: unit tests remain valuable for deterministic helpers. Allowing force-push to main without reviews is wrong: unprotected main branches undermine governance. Skipping infrastructure validation entirely is wrong: infrastructure validation prevents environment drift that breaks agents in subtle ways. Store evaluation thresholds as code-reviewed configuration.