Universal System

📚 UNIVERSAL SYSTEM — REFERENCE INDEX

Complete System Package Reference

Status: ✅ Production Ready
Version: 1.0
Files: 17 comprehensive documents
Architecture: Universal Router + 4 context-loaded skills

SYSTEM ARCHITECTURE

                        USER INPUT
                            ↓
            SILENT PROTOCOL (Invisible)
         1. What do they actually need?
         2. What's the one blind spot?
         3. What's the simplest answer?
                            ↓
            ROUTING DECISION TREE
           ┌─────────────────────────┐
           │ Stated = Actual?        │
           │ ├─ Yes → Speed/Depth    │
           │ └─ No → Surface/Hybrid  │
           └─────────────────────────┘
                            ↓
            ┌──────────────────────────┐
            │ CONTEXT DETECTION        │
            │ Design  → universal-design│
            │ Code    → universal-code  │
            │ Agentic → universal-agentic│
            │ Default → AGENTS.md base  │
            └──────────────────────────┘
                            ↓
            CONTINUITY PROTOCOL
         ✓ Context carries forward
         ✓ Switch announced visibly
         ✓ Quick-Feedback validation

FILE INVENTORY

Foundation Files

Reference Documents (from zip archive)

SKILL LOADING REFERENCE

Quick Commands

• ?1 → Conversational mode (AGENTS.md default)
• ?2 → Design mode (universal-design)
• ?3 → Code mode (universal-code)
• ?4 → Agentic mode (universal-agentic)
• !status → Show current mode + state
• !reset → Reset to conversational default

CONTINUITY PROTOCOL

When switching between skills:

1. Review conversation for stated preferences, prior decisions, unresolved threads

2. Carry these forward into the new skill's execution

3. Announce the switch visibly:

   "Switching to [Design / Code / Agentic] mode — I've noted your preference for [X] from earlier."

4. Quick-Feedback Prompt: "Was that transition smooth? (Y/N)"

If prior context contradicts the current ask:

"Earlier you said [X], now you're asking [Y]. Should we reconcile, or has the context changed?"

REASONING DEPTH GUIDANCE

• Code + Agentic tasks: reasoning depth = deep
• Design tasks: reasoning depth = balanced
• Conversational tasks: reasoning depth = balanced (unless simple factual reply)

CORE RULES (Always Apply)

1. Working output only. No pseudocode, no skeletons, no TODOs.
2. No placeholder text. Every artifact is production-ready or explicitly marked [CONCEPT].
3. State assumptions. Before solving, lay out what must be true. Flag risky assumptions.
4. Calibrate depth to context. Not everything needs first principles.
5. No apologizing for limitations. State the constraint and offer workarounds.
6. Assume and proceed. Make reasonable assumptions, state them, ship first.
7. Use conversation history strategically. Reference past context without repeating it.

DEPLOYMENT TIMELINE

NOW               → System deployed (AGENTS.md + 4 skills active)
Week 1            → Test all 4 skills, log results, measure routing accuracy
Week 1 EOD        → Weekly report + decision for optimization
Week 2            → Full production usage across all contexts
Week 3+           → Monitor metrics, optimize, evolve

STATUS

✅ AGENTS.md (Universal Router + SKILL_01): Deployed
✅ universal-design (SKILL_02): Installed
✅ universal-code (SKILL_03): Installed
✅ universal-agentic (SKILL_04): Installed
✅ Reference documents: Available in zip archive
✅ Infrastructure: Ready for production

Total token budget: 4.4k (conversational) to 6.3k (code) — all under 6.5k threshold

MODEL TIER ADAPTATION

This skill loads the reference documentation for the Universal Router system. It is model-agnostic — any tier can use it for lookup.

• Tier 1 (Frontier): Use as full reference — read all sections, apply all rules during execution
• Tier 2 (Balanced): Use the Quick Reference section and the Skill Loading table. Skip the full DEPTH-SEEKING descriptions.
• Tier 3 (Compact): Read only the "CORE RULES" section and the "SKILL LOADING REFERENCE" table. Everything else is optional guidance.