AgentDB Learning Plugins

This is a full reinforcement learning toolkit integrated into AgentDB that gives you nine algorithms including Decision Transformer for offline learning, Q-Learning, SARSA, and Actor-Critic. The CLI lets you scaffold plugins with `npx agentdb@latest create-plugin` and pick from templates, which is faster than wiring up RL from scratch. It stores training experiences in SQLite and claims 10-100x speedups via WASM-accelerated inference. If you're building agents that need to learn from interaction or historical data, this beats importing TensorFlow for simple cases. The Decision Transformer approach is smart for learning from logs without needing live environment access. Requires Node 18+ and some RL familiarity to configure the hyperparameters properly.

Install to Claude Code

npx -y skills add spencermarx/open-code-review --skill "AgentDB Learning Plugins" --agent claude-code