ReasoningBank — адаптивное обучение AI-агентов на опыте

★ 8.1 · vibe-coding

reasoningbank-intelligence is a Claude Code skill that integrates ReasoningBank's adaptive learning system into AI agents, enabling them to record task outcomes, recognize patterns, and receive strategy recommendations based on accumulated experience. It requires agentic-flow v3.0.0-alpha.1+ and AgentDB v3.0.0-alpha.10+ for persistent storage with optional vector search. Core operations include `recordExperience`, `recommendStrategy`, `learnPattern`, `matchPatterns`, `compareStrategies`, and `enableAutoLearning` — configurable with confidence thresholds and model update frequency. Advanced capabilities cover meta-learning via `metaLearn` and cross-domain knowledge transfer via `transferKnowledge`, such as applying JavaScript code-review patterns to TypeScript. The skill is built for developers creating self-improving agents, optimizing repetitive workflows, or implementing meta-cognitive reasoning pipelines.