∞
Duo-node Logo
|
RJS-Software Labs Logo

System Layout: File Hierarchy & Authority

The exact file hierarchy Omnipotent expects you to follow for instructions, planning, and telemetry.

File Structure

Use this as the canonical mental model for where “law”, “plan”, and “truth” live. When something conflicts, follow the authority order defined by the active Realm.

~/ (Home)
 ├── AI_RULES.md                    → Pointer to ~/.AI_RULES/.AI_RULES.md
 ├── .AI_RULES/
 │   └── .AI_RULES.md               → Multirealm baseline + Realm OS Engine Template
 │       └── ORCHESTRATION_TEMPLATES/ → Reference templates for new Realms
 ├── .claude/
 │   ├── CLAUDE.md                   → Global user instructions (points to ~/AI_RULES.md)
 │   ├── settings.json               → Permissions, hooks (SessionEnd→Graphiti), plugins
 │   ├── settings.local.json         → MCP servers (graphiti, supabase), extra permissions
 │   ├── plans/                      → Claude Code plan files (per-session)
 │   └── projects/                   → Per-project memory (MEMORY.md)
 ├── .config/graphiti/
 │   ├── ingest.py                   → Auto-ingestion engine (SessionEnd hook target)
 │   └── learning-objectives.yaml    → Objective patterns, priority 1-10
 │
 └── Projects/rjs-software-lab/
     └── omnipotent/                 → Active Realm
         ├── AI_RULES.md             → Omnipotent Constitution (overrides Multirealm)
         ├── OMNIPOTENT_PLAN.md      → Canon plan (authority #2)
         ├── OMNIPOTENT_STATUS.md    → Ground truth (authority #1)
         ├── OMNIPOTENT_ABSTRACT.md  → Execution spec
         ├── GEMINI.md               → Gemini agent instructions
         ├── CODEX.md                → Codex (OpenAI) instructions
         ├── .realm/
         │   ├── agents.registry.yaml
         │   ├── orchestration.policy.yaml
         │   ├── telemetry/swarm_state.json
         │   ├── telemetry/heartbeat.log
         │   ├── scripts/switch-to-*.sh
         │   └── security/integrity_manifest.json
         ├── kernel/CLAUDE.md         → Kernel-specific Claude instructions
         ├── ai/shadow-nexus/CLAUDE.md → Shadow Nexus Claude instructions
         └── tools/omnipotent-cli/CLAUDE.md → CLI-specific Claude instructions

How To Use This

If you are changing behavior

  • 1. Start with the active Realm's constitution: omnipotent/AI_RULES.md
  • 2. Use OMNIPOTENT_STATUS.md as ground truth.
  • 3. Keep plans aligned with reality: OMNIPOTENT_PLAN.md.

If you are verifying execution

  • Check the blackboard: .realm/telemetry/swarm_state.json
  • Check integrity: .realm/security/integrity_manifest.json
  • Use the switch scripts when multi-controller: .realm/scripts/switch-to-*.sh