Replication Package: On the Impact of Requirement Smells in LLM-Based Code Generation

Context: Software requirements are typically incorporated into prompts used in LLM-assisted software development. Recent work has shown that requirement smells can affect automated traceability between requirements and code, but empirical evidence on their effects in code generation remains limited. Method: To address this gap, we conceptually replicated and extended a prior study on automated traceability by evaluating how smelly requirements affect functional correctness of LLM-generated code. Using a benchmark consisting of requirements and corresponding system tests for four game applications, we progressively introduced semantic, syntactic, and lexical smells into otherwise clear requirements and analyzed their influence on generated implementations. Results: Our findings show that requirement smells do not affect LLM-based code generation equally. Increasing smell density generally reduced functional correctness, while semantic smells were frequently associated with noticeable degradation patterns, particularly in tasks involving more complex behavior. Compared with prior findings on automated traceability, our results suggest that code generation tasks may be more sensitive to requirement quality issues. Contribution: This work contributes empirical evidence that the impact of requirement smells may depend on the characteristics of the underlying software engineering task. Our findings motivate further investigation into task-sensitive quality effects and reinforce the relevance of requirements quality knowledge in LLM-assisted software development.