Upgrades Claude Code skills to the v5 Hybrid format using decision theory frameworks and modal logic verification.
Skill Upgrader is a sophisticated meta-skill designed to evolve standard SKILL.md files into high-fidelity v5 Hybrid versions. By employing four parallel sub-agents powered by Ragie-backed RAG (referencing foundational texts on planning algorithms, reinforcement learning, and modal logic), it formalizes agent behaviors into Markov Decision Processes (MDP). This process ensures that skills are not just simple instructions, but mathematically sound policies with clearly defined state-action spaces, reward structures, and rigorous safety verifications for mission-critical automation.
Key Features
01Logical constraint verification including temporal and epistemic properties
020 GitHub stars
03Deep RAG integration with authoritative texts on RL and logic
04Automated conversion to Decision Theory v5 Hybrid format
05Formal state-action space modeling and MDP structure generation
06Multi-agent orchestration using four specialized logical roles
Use Cases
01Refining simple prompt-based skills into robust, structured logical agents
02Adding formal safety and liveness constraints to developer workflows
03Standardizing custom skill libraries to a unified, high-performance architecture
