Prompt-Driven Code Summarization: A Systematic Literature Review

This collection of recent arXiv papers surveys and advances LLM-driven software engineering techniques and tooling with clear security-relevant implications. Key contributions include: (1) systematic review of prompt engineering for LLM-based code summarization; (2) Chain-of-Thought (CoT) prompting to significantly improve control-flow deobfuscation and semantic recovery (dual-use for malware analysis and reverse engineering); (3) GraphQLify, an automated static-analysis-based REST→GraphQL migration tool that embeds servers to reduce latency but alters API surface and invocation paths; (4) LLM