#llm-agent

GenericAgent is a minimalist, self-evolving autonomous agent framework with only ~3K lines of core code. Utilizing 9 atomic tools and a ~100-line Agent Loop, it grants any LLM system-level control over local computers, covering browsers, terminals, filesystems, keyboard/mouse input, screen vision, and mobile devices. Its design philosophy is to evolve skills rather than pre-loading them, automatically crystallizing execution paths into reusable skills after each task, fostering a personal skill tree that grows with use, all while being highly token-efficient.