memory-ecosystemlisted

# Memory Ecosystem — Session Protocol > This skill defines HOW Claude should use the 4 memory layers during every session. > Run this protocol at session start and before major code changes. ## 4 Memory Layers (Always Available) | Layer | Tool | Question It Answers | When | |-------|------|-------------------|------| | **L1** | Graphify (knowledge_graph.json) | "What is this project? Business logic?" | Session start | | **L2** | GitNexus MCP (impact/context/query) | "What breaks if I change X?" | Before code changes | | **L3** | Long-term memory (semantic search) | "What did we discuss/decide before?" | When recalling history | | **L4** | NotebookLM MCP (notebook_query) | "What do external sources say?" | Research tasks | ## Session Start Protocol ``` 1. READ knowledge_graph.json (Graphify) → Understand clusters, critical_rules, architecture → ~2,600 tokens, instant orientation 2. READ SPRINT.md + learnings.md (Dev Wiki) → Current tasks, recent lessons 3. IF user mentions past decision → long-term memory search → "search memory for [topic]" ``` ## Before Code Change Protocol ``` IF project has .gitnexus/ folder: 1. impact({target: "[file/symbol]", direction: "both", minConfidence: 0.7}) 2. Review blast radius 3. THEN make the change IF project has NO .gitnexus/: → Use Graphify cluster connections + grep for references → RECOMMEND: "This project would benefit from GitNexus. Run: npx gitnexus analyze" ``` ## Research Protocol ``` IF user asks a