AI Agent's Four Knowledge Sources: Skills, MCP, RAG, Memory

Based on an IBM Technology video, AI Agents rely on four distinct knowledge sources. Skills are structured instructions for specific tasks, using progressive disclosure to load only when needed, but they cannot access external systems. MCP (Model Context Protocol) provides a standardized interface for agents to connect to external tools and data sources, enabling actions like querying logs. RAG (Retrieval-Augmented Generation) pulls relevant documents from a vector database on demand, using human-curated explicit knowledge. Memory stores the agent's own experiential knowledge from past interactions, allowing it to recall solutions from previous similar problems. A decision rule helps choose: written knowledge → RAG, experience → Memory, repeatable process → Skill, real-world action → MCP.