SkillWrapper: Generative Predicate Invention for Task-level Planning

SkillWrapper addresses the 'symbol grounding' problem by using generative AI to invent the symbolic predicates necessary for classical planning (PDDL). While technically sound and addressing a major bottleneck in robotics (the manual authoring of planning domains), the project's defensibility is low (3/10) because it functions primarily as a research artifact. The 11 forks against 0 stars indicate academic interest and replication efforts, but it lacks the community or proprietary data required for a moat. Frontier labs (Google DeepMind, OpenAI) are aggressively pursuing the 'LLM-as-Planner' and 'LLM-for-Abstraction' space through projects like SayCan, PaLM-E, and RT-2. The specific technique of predicate invention is likely to be absorbed into broader 'World Model' or 'Agentic' architectures within the next 12-24 months. Its primary value today is as a modular algorithm for researchers building hybrid symbolic-neural systems, but it faces high displacement risk as foundation models gain better spatial and logical reasoning natively.