LLM Fuzz CI By Louis Abraham and Nicolas Devatine Oct 5, 2026 5 min read A GitHub Action that lets a coding agent write adversarial inputs for your tests, and reports only the ones that fail.