Artifact-driven Alignment of Project Planning and DevSecOps Pipelines in Microservice Systems

Abstract

This paper proposes AAPD, an artifact-driven framework to improve alignment between project planning and DevSecOps pipeline execution through bidirectional traceability and feedback propagation. AAPD organizes software delivery into operational and mapping layers connected through validated cross-layer references among planning, abstraction, architectural, execution, and pipeline artifacts. A unified artifact meta-model, reference validation rules, and trace structures support impact analysis, provenance tracking, and traceability across development activities. Pipeline execution feedback is propagated to planning artifacts to support iterative planning refinement across development rounds. Human-AI collaboration is integrated through provenance, confidence evaluation, and supplementary review mechanisms for controlled cross-layer referencing. We evaluated AAPD on an e-wallet microservice case study across six change scenarios and ten development rounds, using manual Jira-based reconstruction and context-free LLM mapping as baselines. AAPD outperformed both baselines in cross-layer reference accuracy and achieved 87.5% end-to-end trace completion, 100% feedback-loop effectiveness, and a 59.4% average reduction in remediation scope. These results demonstrate AAPD’s capacity to align planning artifacts with DevSecOps execution while supporting remediation scoping and feedback-driven planning refinement.