oraclelisted

# Oracle — Project Knowledge Query ## Context: "$ARGUMENTS" ## Overview The oracle is the **reader** of the project's knowledge vault. It answers questions by traversing MOCs (Maps of Content), design docs, postmortems, GitHub issues, commit history, and code — returning traced answers with source citations. **Relationship to other tools:** - `/obsidian` **writes** to the vault (notes, MOC updates) - `/oracle` **reads** from the vault (queries, traversal, connection discovery) - `code-finder` agent searches **code** (functions, classes, patterns) ## Process ### 1. Determine Scope & Complexity ```bash PROJ_ROOT=$(git rev-parse --show-toplevel 2>/dev/null || pwd) ``` Assess the question: - **Simple lookup** (list MOCs, find a doc by name, check an issue) → handle inline below - **Deep traversal** (trace evolution of a concept, connect issues to code to design docs) → spawn the `oracle` agent via Task tool For deep traversal, delegate: ``` Task(subagent_type="oracle", prompt="<the user's question with project root context>") ``` ### 2. Locate the Knowledge Vault Scan for the standard project layout (check in order): 1. `$PROJ_ROOT/private/claude/` — primary vault 2. `$PROJ_ROOT/private/claude/_maps/` — MOC files (start here) 3. `$PROJ_ROOT/private/claude/_oracle/` — oracle metadata (manifest, concepts index) 3b. `$PROJ_ROOT/private/claude/notes/` — working notes (`ideas/`, `bugs/`, `meta/`, and any project-specific subdirs) 4. `$PROJ_ROOT/CLAUDE.md`, `README.md`, `CHA