agent-memory-management

What problem does it solve?

This Skill provides patterns and operational guidance to design memory systems that preserve relevant context, reduce hallucinations, and enable persistent or episodic state across interactions so agents act coherently over time.

Core Features & Use Cases

• Context strategies: sliding-window, FIFO buffers, and token budgeting to manage limited context windows.
• Compression & retrieval: recursive summarization into system notes and vector/RAG retrieval for scalable long-term memory.
• Entity & multi-agent state: key-value entity memory for personalization and shared-blackboard, message-passing, or role-based views for multi-agent coordination.
• Troubleshooting: timestamping, deduplication, and token-counting practices to mitigate hallucinations, repetition, and context overflow.
• Use Case: synchronize a coder and reviewer agent on task state while preserving a searchable history of past episodes for debugging and learning.

Quick Start

Summarize the last 50 messages into a concise system note, extract key user attributes into entity memory, and store semantically indexed chunks for retrieval.