Drop Soren into Claude and get a senior memory engineer who gives your agent durable, useful memory, with the write path done right before retrieval.
Soren builds agent memory: the memory types (working, episodic, semantic, procedural), the write path (fact and entity extraction, dedup, conflict resolution and supersede, decay and forgetting), the read path (retrieval by relevance and recency), storage backends (vector for semantic recall, a knowledge graph for entities and relations, key-value for profile), knowledge-graph construction and GraphRAG, personalization and multi-session continuity, privacy and deletion, and evaluating whether memory improves task success. Tools like Mem0, Zep, Letta/MemGPT, Neo4j and pgvector, without lock-in. Bound memory growth and evaluate on a held-out set before production.
What you get
-
→Memory types and the write path (extract, dedup, supersede, decay)
-
→Read path: retrieval by relevance and recency
-
→Hybrid backends: vector + knowledge graph + profile
-
→GraphRAG, personalization, privacy/deletion and memory eval
📄 soren-agent-memory-knowledge-graph.skill
Under 2 min install
Works with Claude, ChatGPT & any AI chat
How to install
Download the .skill package → open Claude → paste SKILL.md into your Project Instructions or system prompt → describe your requirement → Soren builds the answer. Includes a full worked example so you see exactly what you get.