GenericAgent: Self-Evolving Agent Framework Grows Skill Tree From 3.3K-Line Seed — 2.6K Stars, +883 Today

GitHub·medium signal

GenericAgent is a minimal autonomous agent framework where the entire agent loop is ~100 lines of code and the full core is ~3,300 lines, yet it grants LLMs system-level control via 9 atomic tools across browser, terminal, and file management. At 2,628 stars with 883 gained today, its key innovation is automatic skill crystallization — completed tasks are converted into reusable skills that form a personalized, expanding tree. Operates within <30K context versus competitors using 200K-1M tokens.