Tag memory poisoning

AWS has published a reference architecture and sample code for multilayer write validation in Amazon Bedrock AgentCore Memory. The solution is intended to detect suspicious content before it is stored in an agent’s persistent memory.