Diagrams

Diagrams — Module 0.2: The Ecosystem

01

Diagram 1 — The Seven-Category Taxonomy

Mermaid (flowchart)
flowchart TB
    H[Any harness]
    H --> C1[Terminal harness<br/>CLI, single-agent, code-focused]
    H --> C2[Platform harness<br/>behind a messaging gateway]
    H --> C3[Meta-harness<br/>plugin layer over a base]
    H --> C4[Orchestration framework<br/>graph or role-based]
    H --> C5[Hardened fork<br/>governance-first fork]
    H --> C6[Academic harness<br/>inspectable, reproducible]
    H --> C7[SDK / runtime<br/>primitives to build your own]

    C1 --> I1[Interesting in:<br/>loop + tool surface]
    C2 --> I2[Interesting in:<br/>cross-channel trust architecture]
    C3 --> I3[Interesting in:<br/>routing + subagent hierarchy]
    C4 --> I4[Interesting in:<br/>state-machine semantics]
    C5 --> I5[Interesting in:<br/>the governance layer ONLY]
    C6 --> I6[Interesting in:<br/>inspectability + reproducibility]
    C7 --> I7[Interesting in:<br/>loop/tool/sandbox abstraction boundaries]

    style C1 fill:#1a1a2e,stroke:#4a4a8a,color:#a0a0d0
    style C2 fill:#1a1a2e,stroke:#4a4a8a,color:#a0a0d0
    style C3 fill:#1a1a2e,stroke:#4a4a8a,color:#a0a0d0
    style C4 fill:#1a1a2e,stroke:#4a4a8a,color:#a0a0d0
    style C5 fill:#2d1515,stroke:#8a2020,color:#d0a0a0
    style C6 fill:#1a1a2e,stroke:#4a4a8a,color:#a0a0d0
    style C7 fill:#1a1a2e,stroke:#4a4a8a,color:#a0a0d0
    style I5 fill:#2d1515,stroke:#8a2020,color:#d0a0a0
02

Diagram 2 — The Lineage Map

Mermaid (graph with fork/competitor edges)
flowchart LR
    AIDER[Aider<br/>2023 · git-first<br/>Polyglot Benchmark]
    PI[Pi<br/>2024 · ~1.2k LOC<br/>4 tools]
    OCODE[OpenCode<br/>2025 · client/server]
    OCLAW[OpenClaw<br/>2025 · 368k stars<br/>40+ channels]
    CC[Claude Code<br/>Anthropic · closed]

    HERMES[Hermes<br/>Feb 2026<br/>layered memory]
    NEMO[NemoClaw<br/>NVIDIA · Apr 2026<br/>governance fork]
    SCOUT[Microsoft Scout<br/>Jun 2026<br/>Entra identity fork]
    OMO[oh-my-opencode<br/>meta-harness<br/>Sisyphus+Prometheus+Atlas]
    OPI[oh-my-pi<br/>extends Pi]

    AIDER -.->|influenced| PI
    CC ==>|clean-room rewrite| OCLAWCLI[Claw Code / OpenClaw CLI]
    OCLAW -->|hardened fork| NEMO
    OCLAW -->|enterprise fork| SCOUT
    OCLAW -.->|competitor, not fork| HERMES
    HERMES -.->|hermes claw migrate| OCLAW
    OCODE --> OMO
    PI --> OPI
    SDK[OpenAI Agents SDK] -.->|sibling| CODEX[Codex CLI<br/>Rust · Apache 2.0]

    style OCLAW fill:#152d15,stroke:#208a20,color:#a0d0a0
    style HERMES fill:#2d2d00,stroke:#8a8a00,color:#d0d000
    style NEMO fill:#2d1515,stroke:#8a2020,color:#d0a0a0
    style SCOUT fill:#2d1515,stroke:#8a2020,color:#d0a0a0
03

Diagram 3 — OpenClaw vs Hermes: Breadth vs Depth

Mermaid (comparison)
flowchart LR
    subgraph BREADTH["OpenClaw — breadth play"]
        direction TB
        OC1[40+ messaging channels]
        OC2[NVIDIA + Microsoft partnerships]
        OC3[#1 by total usage until May 2026]
        OC4[Architectural debate: cross-channel trust boundaries]
        OC5[-> teaches Module 6 + Course 2 trust model]
    end

    subgraph DEPTH["Hermes — depth play"]
        direction TB
        HE1[Layered persistent memory]
        HE2[Self-evolving skills]
        HE3[Overtook OpenClaw on OpenRouter May 10 2026]
        HE4[Architectural debate: memory poisoning surface]
        HE5[-> teaches Module 4 memory architecture]
    end

    BREADTH <-->|opposite bets, both correct| DEPTH

    style BREADTH fill:#0d2818,stroke:#1e8449,color:#82e0aa
    style DEPTH fill:#2a2a0d,stroke:#8a8a00,color:#d0d000
04

Diagram 4 — The Governance Layer Beneath the Agent

Mermaid (layered)
flowchart TB
    subgraph AGENT["Agent process (reachable by the agent itself)"]
        direction LR
        LOOP[Loop]
        TOOLS[Tools]
        PERMS[Permission flags<br/>system-prompt rules]
    end

    subgraph GOV["Governance layer (OUTSIDE the agent's reach)"]
        direction LR
        NEMO_GUARD[NeMo Guardrails<br/>evaluates every model call]
        OPENSHELL[OpenShell sandboxes<br/>agent never touches directly]
        ENTRA[Entra identity<br/>every action attributable]
    end

    WORLD[(World: filesystem, network, channels)]

    AGENT --> GOV --> WORLD

    style AGENT fill:#15152d,stroke:#20208a,color:#a0a0d0
    style GOV fill:#2d1515,stroke:#8a2020,color:#d0a0a0
    style WORLD fill:#1a1a2e,stroke:#4a4a8a,color:#a0a0d0
05

Diagram 5 — Study Priority Heatmap

draw.io XML + summary table
06

Diagram 6 — Meta-Harness Routing (n8n)

n8n importable workflow JSON
Meta-Harness Routing — Module 0.2
Task Intake webhook Normalize Task set Classify Task AI:Model Call Route to Sub-Harness switch Recon Sub-Harness executeCommand Web Review Sub-Harness executeCommand CTF Solver Sub-Harness executeCommand Aggregate Findings + Cost set Report respondToWebhook
Import this workflow: n8n → Workflows → Import from File. Hover nodes for details.