Agent Memory Systems

The fundamental challenge in agent systems isn't storing information, it's retrieving the right memory at the right time. This skill treats memory as a retrieval problem first: how you chunk documents, which vector store you choose, and how you filter by metadata determine whether your agent appears intelligent or amnesiac. It covers the full stack from short-term context windows to long-term vector stores, with practical patterns for chunking strategies and contextual retrieval. The sharp edges table is unusually honest about what actually breaks in production, like embedding model drift and memory conflicts. If you're building agents that need to maintain context across sessions or access large knowledge bases, the retrieval strategies here will save you from the "why did it forget that?" debugging sessions.