仓库原始内容
按源仓库内容呈现，保留标题、案例、代码、表格、链接以及原文引用的演示图片。
/understand-coremind
Query the CoreMind knowledge graph with governance-aware context. This skill understands the GAOS architecture, agent hierarchy, and 10-stage pipeline.
Instructions
- Load the knowledge graph from
CoreMind/.understand-anything/knowledge-graph.json - Parse the query from
$ARGUMENTSand determine query type:
Query Types
Architecture queries ("how does X work", "what is the pipeline", "explain governance"):
- Map the query to relevant layers and nodes in the graph
- Explain using the orchestrator governance model:
- 10-stage pipeline: IntentParser -> PolicyGate -> GoalLedger -> Planner -> DelegationEngine -> AgentCoordinator -> ReflectionLoop -> OutcomeTracker -> WorldState
- 4-tier governance: ALLOW (auto-approve) / REVIEW (human check) / ESCALATE (multi-agent review) / BLOCK (deny)
- 5-layer safety stack: Input sanitization -> Policy gate -> Execution sandbox -> Output validation -> Recovery
Agent queries ("which agents handle X", "what does the CEO agent do"):
- Search the Agent Framework layer (236 nodes)
- Map agent hierarchy: L0 System Core -> L1 Executives (CEO/CTO/CFO/CMO/COO) -> L2+ Specialists
- Include agent capabilities, tools, and delegation patterns
Module queries ("what does the API layer do", "how does knowledge store work"):
- Search the relevant architecture layer
- Include: module purpose, file count, key components, dependencies (from module edges)
- Cross-reference with Memory MCP entities:
mcp__memory__search_nodes("CoreMind:Module:{name}")
Impact queries ("what would break if I change X", "dependencies of core_mind"):
- Find the target node in the graph
- Traverse 1-hop edges (imports, calls, inherits, depends_on)
- Identify cross-layer impact
- Rate risk: LOW (same layer) / MEDIUM (adjacent layers) / HIGH (3+ layers affected)
Tour queries ("give me the tour", "where do I start"):
- Return the 15-step guided tour from the knowledge graph
- Each step: title, node, description, why it matters
CoreMind Architecture Layers (8)
| Layer | Nodes | Key Components |
|---|---|---|
| Orchestration Core | 276 | CoreMind, Planner, PolicyGate, IntentParser, GoalLedger, DelegationEngine |
| Agent Framework | 236 | BaseAgent, executives (CEO/CTO/CFO/CMO/COO), tools, capabilities |
| Integration Layer | 454 | FastAPI routes, middleware, connectors, interfaces, billing |
| Governance & Safety | 146 | GAOS (5-layer safety), security (RBAC, JWT, hash chain, audit) |
| Knowledge & Learning | 127 | RAG pipeline, knowledge store, evaluation benchmarks |
| Data & Persistence | 181 | SQLAlchemy models (15 entity types), repositories |
| Presentation | 436 | Next.js frontend, 5 product surfaces (Forge, Helios, Nova, Command, Studio) |
| Cross-Cutting | 55 | LLM providers (OpenAI, Anthropic, Ollama, Kimi), model router |
Response Format
Always include:
- Direct answer to the query
- Graph context: which nodes/layers are relevant
- Governance implications: does this touch governance/safety layers?
- Suggestion: related areas to explore or commands to run (
/understand-explain,/understand-chat)
