1Episodic, Semantic, and Procedural Memory for AgentsReliable agents need more than a big context window. They need three kinds of memory, each answering a different question and carrying a different trust.Advanced 2The Context Graph: Wiring Repos, Policies, and IdentityMemory is what you store; context is what you show. A context graph is the relationship layer that makes the choice smart enough for security.Advanced 3Memory as an Attack Surface: Poisoning, Staleness, LeakageMemory makes agents reliable, but it is also something to attack. Stored observations can become tomorrow's trusted instructions.Advanced 4The Promotion Pipeline: From Observation to Trusted KnowledgeNever write a raw observation straight into trusted memory. Make it earn its way up, stage by stage, and expire when it no longer holds.Advanced 5Labeled Recall: Information-Flow Control for MemoryGive every memory a label for source, trust, freshness, and owner, and let the label — not the model — decide what that memory may do.Advanced