Researchers have developed ContextSniper, a module for large language model agents that helps them repair code issues in repositories more efficiently. The tool indexes code and action memory at three abstract levels, retrieves relevant candidates with a hybrid ranker, filters out unnecessary information, and returns compact evidence packets. In tests on two benchmark tasks, ContextSniper reduced token use by 51.5% to 38.9%, depending on the task, without sacrificing repair quality. --- Why it matters: This matters because it could help large language model agents that repair code issues in repositories to do so more efficiently and with lower costs, making them more practical for real-world use. Source: https://arxiv.org/abs/2607.01916