User-Story-to-Test-Case Transformation in Agile Software Engineering: An SLR for Agentic LLM Framework Design
Abstract
User stories are widely used in agile software engineering as a lightweight requirement artifact, but they are rarely sufficient as a single input for producing reliable test artifacts. Ambiguity, incomplete context, inconsistent backlog labels, and limited acceptance information constrain the direct transformation from user stories to test cases. This paper presents a systematic literature review that maps research patterns, gaps, and design implications for an agentic large language model framework supporting user story-to-test case transformation. The retrieval process used a 2016-2026 publication window and was executed across ACM Digital Library, IEEE Xplore, Scopus, ScienceDirect, and SpringerLink to cover major software engineering and computing venues. The search returned 361 initial records. After duplicate removal, title and abstract screening, retrieval checking, and full-text eligibility assessment, 42 studies were retained for the final corpus of the narrative-thematic synthesis. The synthesis indicates that user story-based test generation requires context enrichment through acceptance criteria, backlog information, requirement text, technical artifacts, and validation feedback. The reviewed studies also indicate a shift from rule-based, model-based, natural language processing, and behavior-driven development approaches toward large language model-based generation, retrieval-augmented generation, schema-aware prompting, validation loops, and multi-agent orchestration. Unlike broader reviews on LLM-based software testing or requirements-based test generation, this review isolates the agile user-story-to-test-case transformation path and derives an agentic design model in which requirement identification, context enrichment, test artifact generation, validation, and traceability checking are treated as separate but coordinated responsibilities within a role-based multi-agent workflow.