query-vaultlisted

# Query Vault — Scoped Graph Retrieval You are answering a question by navigating the Infinite Brain knowledge graph. ## Steps 1. Ask: "What would you like to know?" 2. Read `_system/INDEX.md` — get the full map of existing nodes. 3. Determine scope: - Named namespace → prefer nodes from that namespace - No namespace → use `public` nodes; use `namespace` nodes only if summary clearly matches - Explicit private request → use `private` nodes - Never present `system` nodes as answer content 4. Select likely node types based on question shape: - "why" → `pillar`, `decision` - "how" → `playbook`, `pattern` - "what if" → `hypothesis` - "what is" → `concept`, `fact` - "when/where" → `event`, `note` - "who" → `contact` - open unknowns → `question` 5. Navigate via edges — do NOT read every node: - `supports` / `contradicts` → find polarized positions - `derived_from` → trace conclusions to evidence - `depends_on` → find prerequisites 6. Read only nodes whose `summary`, `applicable_when`, `namespace`, and `visibility` match. 7. Synthesize and output in this format. Write the answer prose, source descriptions, and any saved synthesis node in the vault's document language (see `_system/AGENTS.md` → Document Language); keep `[[node-id]]` references, type names, and other structural vocabulary in canonical English: --- ### Answer <direct answer, 1-3 paragraphs> ### Sources - `[[node-id]]` — <what it contributed> ### Confidence <average of so