| name | munin-architecture |
| description | Use when discussing "architecture", "tech stack", "system design", "kiến trúc", "thiết kế", "cấu trúc project", or how components fit together. Also use when proposing or discussing technical changes. |
Munin Architecture Skill
What I Do
I surface all architecture decisions, technical choices, and system design context from your project's memory. Before proposing any technical change, I check what's already been decided.
MCP Tools
| Tool | Purpose |
|---|
munin_search_memories | Search architecture memories with natural language |
munin_retrieve_memory | Get specific architecture decision by key |
munin_diff_memory | Compare architecture versions during migrations |
munin_recent_memories | Get recently updated memories for context |
Usage Protocol
Before Any Architecture Discussion
- Call
munin_search_memories with: relevant subsystem + "architecture decision"
- Call
munin_retrieve_memory for the canonical tech stack memory if found
- Present existing architecture context before suggesting changes
Before Proposing Architecture Changes
- Search for the current implementation in memory
- Document the before state (current architecture)
- After the discussion, store the after state with tags
["architecture", "decision"]
Migration Tracking
When a migration happens:
- Store with title:
"Migrate from X to Y — {date}"
- Tags:
["architecture", "migration", "X", "Y"]
- Content: Include version/timeline anchors (e.g., "v1.2.3 → v2.0.0")
Example Query Patterns
"api gateway architecture"
"MongoDB schema decisions"
"frontend state management architecture"
"auth system design"
"database connection pooling"
GraphRAG Integration
Architecture memories feed into the entity graph:
- Nodes: Services (API, DB, Cache), Modules, Libraries
- Edges: "uses", "depends-on", "connects-to", "wraps"
- Search returns both raw memories AND extracted graph relationships
This enables questions like "which service handles authentication?" or "what does the payment service depend on?"
