Agent Memory Systems

A design skill for long-term memory in agent systems. Your agent gets guidance on keeping knowledge across sessions, tracking people and things over time, marking when a fact stops being true, picking graph or vector retrieval, and merging memories so they don't pile up.

It also helps you choose a benchmark to test whether your memory setup actually works. Scratchpads, handoff summaries and token tricks are handled by sister skills in the same pack.

Part of the Agent Skills for Context Engineering pack by Muratcan Koylan.