I've found that a simple markdown knowledgebase (with some useful extensions like semantic search & git context) is all I need to improve the memory of my agent

I've found that a simple markdown knowledgebase (with some useful extensions like semantic search & git context) is all I need to improve the memory of my agents. Even my non-coding agents have a memory repo. Here's my implementation: https://hraness.com/kb https://hraness.com/kb