cognitive-architectures

What problem does it solve?

This Skill helps engineers and researchers move beyond stateless prompt-response interactions by defining modular, stateful cognitive architectures that manage memory, perception, action selection, and planning to produce reliable, goal-directed language agents.

Core Features & Use Cases

• Modular Memory Models: Patterns for working, episodic, semantic, and procedural memory with recommendations for hybrid (keyword + vector) retrieval and memory decay.
• Action Space & Tool Integration: Guidance on defining external and internal actions, token-efficient tool descriptions, and structured outputs (JSON) for tool arguments.
• Decision Loop & Safety: Decision-making templates (ReAct, OODA, Plan-and-Solve), reflection strategies, retry/fallback controls, and mitigation for context overflow and action loops.
• Use Case: Build an autonomous assistant that preserves long-term user preferences, plans multi-step tasks, and safely integrates APIs and knowledge stores.

Quick Start

Design an agent that records the last 10 interactions in episodic memory, summarizes them into semantic memory, and uses a ReAct-style decision loop to select tools and actions.