orama-system

This is the Legacy ὅραμα System

moving to 2.0 on another repo soon

"Technology married with humanities yields solutions that make hearts sing." (Inspired by ultrathink.)

The complete agent methodology for solving impossible problems with elegance.

Architecture Map

ARCHITECTURE.md is the concise routing map for the method, skill-to-graph escalation, evidence boundaries, and the v1-to-v2 separation. It complements the operational guides below; it does not replace them.

Compatible with: Claude Code · Cowork · Clawdbot · MoltBot · OpenClaw · ECC Tools (everything-claude-code) · Cursor · Hermes · Codex · Cline · Kimi · Gemini · Antigravity

What This Is

A production-ready, self-improving agent skill package that synthesizes:

The Amplifier Principle

AI doesn't replace judgment, it amplifies intent: ultrathink gives that intent structure. Clear methodology → clear output. Vague intent → scaled ambiguity.

Quick Start (3 Minutes)

# 1. Install bin/orama-system (Claude Code / Cowork / OpenClaw)
./install.sh

# 2. Activate in Claude
# "ultrathink this"
# "Apply The ὅραμα System to: [your task]"

# 3. Create a task plan
./bin/orama-system/scripts/create_task_plan.sh "Build my feature"

# 4. Verify before marking done
python bin/orama-system/scripts/verify_before_done.py --task "Build my feature"

Repository Structure

orama-system/
│
├── bin/orama-system/               ← Install here for Claude Code / Cowork
│   ├── SKILL.md                    ← Main intelligence layer (<500 lines)
│   ├── references/                 ← Deep-dive documentation
│   │   ├── oramasys-5-stages.md
│   │   ├── core-operational-directives.md
│   │   ├── content-insertion-framework.md
│   │   └── skill-architecture-guide.md
│   ├── scripts/                    ← Automation tools
│   │   ├── verify_before_done.py   ← Pre-completion verification
│   │   ├── capture_lesson.py       ← Self-improvement loop
│   │   └── create_task_plan.sh     ← Task plan generator
│   └── templates/                  ← Reusable task / lesson templates
│
├── bin/                    ← Install here for Clawdbot / OpenClaw
│   ├── agents/                     ← 7 specialized agents
│   │   ├── orchestrator/           ← Coordinates 5-stage process
│   │   ├── context/                ← Stage 1: Context Immersion
│   │   ├── architect/              ← Stage 2: Visionary Architecture
│   │   ├── refiner/                ← Stage 3: Ruthless Refinement
│   │   ├── executor/               ← Stage 4: Masterful Execution (parallelizable)
│   │   ├── verifier/               ← Stage 4.5: Verification Before Done
│   │   └── crystallizer/           ← Stage 5: Crystallize the Vision
│   ├── shared/                     ← Core types + state + messaging
│   ├── config/                     ← Agent registry + routing rules
│   └── mcp_servers/                ← MCP server implementations
│
├── examples/                       ← Real-world usage walkthroughs
├── docs/                           ← Installation, guides, FAQ
├── portal_server.py                ← LAN dashboard for PT, ultrathink, LM Studio, Ollama
├── scripts/                        ← Helper scripts (setup, network, install)
├── tests/                          ← Full test suite (pytest)
└── .github/                        ← CI/CD workflows

Installation

Single-Agent (Most Users)

./install.sh

Multi-Agent Network (Distributed / Parallel)

scripts/install-multi-agent.sh
python bin/mcp_servers/ultrathink_orchestration_server.py

LAN Helpers

# Probe a LAN-friendly host/IP and optionally scan for agents
python scripts/network_autoconfig.py --scan

# Launch the slate-grey portal on port 8002
python portal_server.py

• scripts/network_autoconfig.py detects the preferred local IP and can scan for LM Studio, Ollama, PT, ultrathink, and portal services.
• portal_server.py shows a lightweight LAN dashboard for PT, ultrathink, LM Studio, and Ollama endpoints.

The 5-Stage Process

1. Context Immersion    — Scan git, docs, patterns, constraints. Understand before acting.
2. Visionary Architecture — Design the most elegant solution. Decompose modularly.
3. Ruthless Refinement — Eliminate everything non-essential. Elegance = nothing left to remove.
4. Masterful Execution — Plan → Craft (TDD) → Verify (programmatic, not visual).
5. Crystallize Vision  — Assumptions ledger, simplification story, inevitability argument.

The 6 Operational Directives

Always active, regardless of which stage you're in:

Multi-Agent Architecture

User → Orchestrator
           ↓
    Context Agent           ← Stage 1 (spawns sub-agents)
           ↓
    Architect Agent         ← Stage 2 (spawns module designers)
           ↓
    Refiner Agent           ← Stage 3 (loops until elegance ≥ 0.8)
           ↓
  ┌──────────────────┐
  │Executor Agent × 5│      ← Stage 4 (parallel)
  └──────────────────┘
           ↓
    Verifier Agent          ← Stage 4.5 (blocks until PASS)
           ↓
    Crystallizer Agent      ← Stage 5 (documents + captures lessons)
           ↓
    Result + Lessons

Self-Improvement System

# After any mistake or user correction:
python scripts/capture_lesson.py

# Review before starting work:
python scripts/capture_lesson.py --review

# Analyze your mistake patterns:
python scripts/capture_lesson.py --stats

Lessons compound over time. Mistake rate measurably declines.

ECC Tools Integration

This skill follows the SKILL.md open standard and is compatible with ECC Tools (65+ skills, 16 agents, 102 AgentShield security rules).

# Add to your ECC profile
cp -r bin/orama-system ~/.claude/skills/orama-system

# Then use with any ECC-compatible harness

Success Metrics

Contributing

See CONTRIBUTING.md. Apply ultrathink to your contributions. Run ./verify-package.sh before submitting a PR.

High-value contributions: new examples, additional reference docs, platform integrations, lessons-learned patterns, security improvements.

License

MIT License — use freely in commercial and private projects. See LICENSE.

GitHub Topics

ai-agents claude claude-code skill-md prompt-engineering ultrathink agent-framework ecc-tools openclaw clawdbot mcp self-improving

"The people crazy enough to believe they can change the world are the ones who do." 🚀