Skip to content
View ElmatadorZ's full-sized avatar
🏠
Working from home
🏠
Working from home

Block or report ElmatadorZ

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
ElmatadorZ/README.md

Bunyawat Dechanon · ElmatadorZ

Independent Cognitive Infrastructure Architect

Cognitive operating systems, agent-governance frameworks, and reusable reasoning protocols — as specifications you can check, and systems you can run.

Capability must never outrun accountability.

A system may be trusted only with what it can be held to afterwards.

Models evolve. Protocols endure.


Followers License

Cognitive OS · Multi-Agent Systems · Agent Governance · Epistemology Engineering · Market Structure · Coffee Science

📍 Kamphaeng Phet, Thailand 🌐 Money Atlas · ☕ Alternative Slowbar : Roaster · ▶️ YouTube · 🎵 TikTok


🚀 Latest — Meta-Cognition Agent OS · Apache-2.0 validate

The flagship — every system below, distilled into one file. IQ is thinking. EQ is thinking about feeling. This is thinking about thinking.

Ask any model to reflect and it will reflect, fluently — "let me examine my assumptions" names no assumption, costs nothing, and is indistinguishable from having done the work. Every meta-cognition framework can be satisfied that way, which means none of them was ever enforcing anything.

So this one rests on a single law: a meta-cognitive step that changed nothing did not happen. Name what changed, or report the step skipped. It is the only requirement fluency cannot fake, and every other rule is written as an observable property of an output — so a third party can check a build without trusting it.

One SKILL.md. No tools, no runtime, no dependencies. Runs on Claude, GPT, Gemini, OpenClaw, Hermes, or anything that reads a system prompt. The five levels · Conformance · Wiki, 13 pages · See it fail first

SkynetClaw · THE HOUSE · Apache-2.0 CI

The architecture above, actually running. A council of 14 agents with an institutional memory: it persists every deliberation, grades its own predictions against reality at fixed horizons, tracks which member was right, preserves dissent, and revises what it believes when reality disagrees.

Local-first — Ollama, llama.cpp, or any cloud API. FastAPI + SQLite, no build step. Proven on Ubuntu and Windows × Python 3.10 / 3.11 / 3.12 on every push, because "it works" is a claim about a machine that is not the author's. Install · Wiki, 21 pages


Start here

Fifteen repositories, one system. If you are new, you do not need all of them — you need one of these three, depending on what you came for.

If you want to… Start with Install
make an agent's reasoning checkable — stop invented specifics, meaningless confidence, agents that never refuse Meta-Cognition Agent OS pip install meta-cognition-agent-os
reduce hallucination in one skill — the epistemic base layer everything else inherits First Principle Codex OS copy SKILL.md
run a system, not read a spec — a council with institutional memory that grades its own predictions SkynetClaw git clone · FastAPI + SQLite

Everything else on this page is either a layer one of those three is built on, or a domain application of them. All repositories are tagged, so any of them can be pinned to a version.


What this account is

Most AI work optimizes the model. This ecosystem optimizes the reasoning — cognitive infrastructure that stays reliable when models change, context changes, incentives change, and environments change.

The repositories here are one system, not a dozen projects. They inherit a single epistemological foundation — the First Principle Codex OS — and every domain skill obeys the same rules: separate Known from Inferred from Unknown, run a non-skippable self-critique gate before output, and state confidence honestly.

I don't publish prompts. I publish operating systems for structured reasoning — as SKILL.md protocols that run on any instruction-following model, and as running systems that anyone can install and audit.


Ecosystem architecture

graph TD
    FPCOS["🧠 First Principle Codex OS<br/>— shared epistemology · the CPU —"]
    GP["⚙️ Genesis Protocol<br/>orchestration standard"]
    GGOV["🏛️ Genesis Governance OS<br/>multi-agent governance"]
    GM["🤝 Genesis Mind<br/>multi-agent reasoning"]
    GC["💓 Genesis Consciousness OS<br/>emotion-weighted reasoning"]
    RG["🎯 Reality Grading<br/>claims graded against outcomes"]
    MA["📊 Money Atlas<br/>financial intelligence · SMC"]
    COF["☕ Alternative Coffee Intelligence<br/>seed → cup"]
    GOSB["📐 Genesis OS Blueprint<br/>runnable reference architecture"]
    SC["🏛️ SkynetClaw · THE HOUSE<br/>the architecture, running"]
    MCOS["🧭 Meta-Cognition Agent OS<br/>— the synthesis · one organ from each —"]

    FPCOS --> GP
    GP --> GGOV
    GP --> GM
    GP --> GC
    GP --> RG
    GM --> MA
    GM --> COF
    GP --> GOSB
    GGOV --> GOSB
    GOSB --> SC
    RG --> SC

    FPCOS -.-> MCOS
    GP -.-> MCOS
    GGOV -.-> MCOS
    GM -.-> MCOS
    GC -.-> MCOS
    RG -.-> MCOS
    SC -.-> MCOS

    classDef base fill:#0d1117,stroke:#d4a017,stroke-width:2px,color:#f0f0f0;
    classDef ship fill:#161b22,stroke:#2ea043,stroke-width:2px,color:#f0f0f0;
    classDef flag fill:#0d1117,stroke:#8957e5,stroke-width:3px,color:#f0f0f0;
    classDef node fill:#161b22,stroke:#30363d,stroke-width:1px,color:#e6e6e6;
    class FPCOS base;
    class SC,GOSB ship;
    class MCOS flag;
    class GP,GGOV,GM,GC,RG,MA,COF node;
Loading

The base layer cannot be removed by any skill that inherits it. Domain skills may add reasoning layers; they may not skip the Reality Anchor, the proof standard, or the self-critique gate. That contract is what keeps answers consistent across domains.

The dotted edges run the other way. Meta-Cognition Agent OS is not built on the ecosystem — it is built from it: one operation per ancestor, each earning its place by having caught a real defect in something that ran. Monitoring and the shadow gate from FPCOS, falsification from Genesis Protocol, multi-frame flexibility from Genesis Mind, state-as-data from Consciousness OS, authority limits from Governance OS, EARNED/UNEARNED calibration from Reality Grading, and the self-development loop from SkynetClaw — where it was actually closed, and where its failures were measured.


Systems you can run

Installable, tested, Apache-2.0.

System What it does Apply it to
SkynetClaw · THE HOUSE An institutional-intelligence operating system. 14-agent council, one SQLite institutional memory, recall that returns justified history rather than raw text, a constitution enforced rather than advised, Bayesian calibrated reputation, and predictions graded at 7 / 30 / 90 / 180 days. 56 tools, 267 routes, 606 tests, CI on two operating systems and three Python versions. Running a council that remembers, on your own hardware — and being able to ask it why it believes something.
Genesis OS — Cognitive Agent Architecture Blueprint The reference architecture: a Cognitive Kernel + ABI, a fail-closed Policy Hook Surface, swappable Capability Providers, and a Reality Grading Loop that grades outcomes against evidence — not the model's own claims. Docs, framework-agnostic specs, ADRs, a dependency-free Python reference, and a wiki. Building agent systems whose capability never outruns their accountability — implement the blueprint, or conform your own build to it.
Genesis Reality Grading The discipline that separates a system that learns from one that only sounds like it: stake a falsifiable hypothesis, judge it with a versioned judge, and let the outcome revise the belief. An abstention is recorded as an abstention — never as a convenient zero. Conformance spec with stable requirement IDs. Any agent that makes claims about the future and should be held to them.
Genesis Governance OS Constitutional framework for multi-agent systems: deny-by-default permissions, separation of powers, irreversible actions gated on a human, and an audit trail that is not optional. RFC-2119 conformance suite + wiki. Deciding what an autonomous system is allowed to do, and proving afterwards what it did.

Reasoning protocols

Model-agnostic SKILL.md systems — the cognition, not the runtime.

System What it does Apply it to
Meta-Cognition Agent OS Apache-2.0 The flagship, and the synthesis of everything below it. A control layer that governs how an agent thinks rather than what it concludes: five operations, five levels read from the tell in the output rather than self-declared, a non-skippable shadow gate that may refuse, confidence carrying EARNED or UNEARNED, and a promotion rule — three cited instances up, one counter-instance down, nothing deleted, dormancy made visible. Conformance IDs MC-0MC-9, each written as an observable property of an output. Any agent whose confidence you cannot currently check — and any framework you suspect is being satisfied by writing as if it had been followed.
First Principle Codex OS Apache-2.0 The anti-hallucination base layer every other skill inherits: Known / Inferred / Unknown separation, a 10-point proof standard, and a mandatory self-critique gate before any output. Any task where being wrong is expensive — research synthesis, due diligence, claims that must hold up.
Genesis Protocol Apache-2.0 OS-level cognitive standard for strategy-capable AI: falsification-first reasoning, refusal integrity, multi-horizon foresight. Giving an agent the discipline to refuse, to flag risk, and to reason over long horizons.
Genesis Mind Multi-agent reasoning OS: specialist coordination, consensus, meta-cognition, recursive self-evaluation. Built on FPCOS. Decisions under uncertainty that need more than one viewpoint — strategy, system design, "what am I missing."
Genesis Consciousness OS Apache-2.0 Experimental architecture treating emotional state as a data layer — detecting it from input and reweighting which reasoning agents lead. Research into affect-aware reasoning; agents that adapt tone and priority to context.

Domain applications

System What it does Apply it to
Money Atlas Apache-2.0 Financial intelligence for markets, macro, and geopolitics. Genesis Protocol reasoning + a Smart Money Concepts layer → structured scenarios with explicit entry/exit zones and stated uncertainty. Market-structure reads, macro framing, trade-thesis stress-testing — not signal-following.
Alternative Coffee Intelligence Specialty-coffee reasoning seed → cup: roast analysis, disease diagnosis, extraction science, brew troubleshooting, business decisions. Roasters and cafés — diagnosing roast and brew problems, sensory analysis, operational calls.

Positions

Four claims this work rests on. Each is stated so that a reader can disagree with it, and each has a repository where it is implemented rather than argued.

The claim Where it is enforced rather than asserted
A meta-cognitive step that changed nothing did not happen. Every framework in this space can be satisfied by writing as if it had been followed — so the only requirement worth having is one that fluency cannot fake. Meta-Cognition Agent OS · MC-0
A provider's assertion is not evidence. A prediction counts only once a versioned judge has graded it against an outcome, and an abstention is recorded as an abstention — never as a convenient zero. Genesis Reality Grading
Authority must be structured before it is granted. Deny-by-default, separation of powers, and irreversible actions gated on a human — an autonomous system's permissions are a constitutional question, not a configuration one. Genesis Governance OS
A named gap beats a confident guess. Separating Known from Inferred from Unknown — and refusing to let an unsourceable claim reach the recommendation — is what makes the gap visible instead of quietly filled. First Principle Codex OS

The point of putting them in code is that none of them can be settled by agreeing with me. A specification whose requirements are observable properties of an output can be checked by someone who does not trust its author — which is the only kind of claim worth publishing.


Operating principles

1 · Reality before narrative. Evidence precedes interpretation. 2 · Falsification before assertion. A claim is accepted only after surviving an active attempt to break it. 3 · Known / Inferred / Unknown. Every analysis labels observed facts, inferred conclusions, and unknown variables — separately. 4 · Self-critique is mandatory. Every system challenges its own output before publication. 5 · Explicit uncertainty. Every conclusion states its confidence level, its failure boundaries, and what would reverse it. 6 · Capability must never outrun accountability. A system may only be trusted with what it can be held to afterwards.


Cognition / execution separation

  Reasoning Layer  →  FPCOS · Genesis Protocol · SKILL.md systems   (invariant)
        │
        ▼
  Execution Layer  →  Python · APIs · Tools · Agents                (interchangeable)
        │
        ▼
  Applications     →  Finance · Logistics · AI · Coffee · Research

The skill is the cognitive brain. The runtime is the body. Never merged — always layered. The reasoning lives in the SKILL.md; the runtime only executes. This is why a skill can move between models without rewriting the logic — and why SkynetClaw can swap Ollama for a cloud API without touching how it thinks.


FAQ

Is this a prompt collection? No. It is a set of cognitive operating systems and reusable reasoning protocols — and, in SkynetClaw's case, a running system with a test suite and continuous integration.

Are these repositories independent projects? No. All of them inherit the same epistemological foundation through FPCOS.

Does it work across different AI models? Yes. The architecture is intentionally model-agnostic. Models are interchangeable; protocols remain invariant.

What problem does this ecosystem solve? Reasoning drift — the way most AI gives different answers as models, context, and framing change. This work makes reasoning stable across all of them, and makes the claims checkable afterwards.


License & citation

Apache-2.0 across the published ecosystem.

The earlier Open Cognitive License was replaced deliberately. A licence that is not OSI-approved — and one carrying a revenue-share clause in particular — is rejected automatically by most corporate open-source review processes, whatever its merits. Apache-2.0 removes that barrier while keeping what actually mattered: §4 requires attribution and preservation of the NOTICE file, and §6 grants no trademark rights, so the names stay mine and no adopter may imply endorsement.

  • Attribution: Built on FPCOS by Bunyawat Dechanon (ElmatadorZ) — carried in each NOTICE.
  • Patent grant and termination are Apache-2.0 standard, in both directions.
  • One exception, on purpose: Alternative Coffee Intelligence is CC0-1.0 — public domain, no attribution asked. Farming knowledge should belong to the farmers.

Canonical citation: Bunyawat Dechanon (ElmatadorZ). First Principle Codex OS (FPCOS): Cognitive Operating System Architecture for AI Reasoning and Agent Governance.


"The best system is one that questions itself — and still functions."

"ระบบที่ดีที่สุดคือระบบที่ตั้งคำถามกับตัวเองได้ — และยังทำงานได้ต่อ"

Bunyawat Dechanon · ElmatadorZ Founder — First Principle Codex OS · Genesis Protocol · SkynetClaw · Money Atlas · Alternative Slowbar : Roaster

Pinned Loading

  1. skynetclaw skynetclaw Public

    An institutional-intelligence operating system: a council of 14 agents that remembers every deliberation, grades its own predictions against reality, and revises what it believes. Runs on your mach…

    Python 15 6

  2. MoneyAtlas-ClaudeSkill-Agent MoneyAtlas-ClaudeSkill-Agent Public

    Money Atlas Skill.md (Claude Skill) For AI Agent A Genesis Protocol-powered intelligence system for financial markets, macroeconomics, and geopolitics. Core Systems - Genesis Protocol (First Princi…

    Python 53 13

  3. FirstPrincipleCodex-OS-Skill FirstPrincipleCodex-OS-Skill Public

    Reduce hallucinations through first-principles reasoning, verification, self-critique, and explicit uncertainty for AI agents.

    Python 35 6

  4. Genesis-Mind-ClaudeSkill-Agent Genesis-Mind-ClaudeSkill-Agent Public

    Cognitive Operating System for AI Agent (Skill.md) Self-thinking Multi-agent reasoning Meta-cognition Self-evolution

    Python 15 3

  5. genesis-os-blueprint genesis-os-blueprint Public

    Architecture blueprint for accountable AI agents — a cognitive kernel, fail-closed policy hooks, swappable model providers, and evidence-graded outcomes.

    Python 2 1

  6. genesis-governance-os genesis-governance-os Public

    Genesis Governance OS The Operating System for Multi-Agent AI Inspired by Political Science, built to coordinate intelligent agents at scale.

    Python 7