Jinqiu Yang

IssueTrojanBench: Benchmarking AI Coding Agents Against Malicious Issue Requests

Ankur Singh, Jinqiu Yang, Tse-Hsun Chen

AI coding agents powered by LLMs are increasingly integrated into real-world software development, where they generate, edit, and execute code with autonomous access to local files and tools. Coding agents inherit security risks from both the LLM backbone, where adversarial promp…Read more