SkillSmith Framework for Local Agent Skill Construction

SkillSmith enables small language models (SLMs) to match cloud-based agent performance by automatically constructing and evolving non-parametric skills. This framework uses skill-based context to supply environment rules and operational procedures, reducing reliance on large-scale backbone models for multi-step tasks.