MSCP Agent Cognition Level Series¶

Status: 🔬 Experimental - All documents in this series describe designs that emerged from prototyping and testing. They are not finalized specifications.

Overview¶

The Minimal Self-Consciousness Protocol (MSCP) defines a six-level taxonomy of agent cognition, from simple tool-calling agents to the theoretical boundary of artificial general intelligence. Each level document includes architecture diagrams, pseudocode, and safety analysis based on what we've explored so far.

%%{init: {'theme': 'base', 'themeVariables': {'primaryColor': '#0078D4', 'primaryTextColor': '#003D6B', 'primaryBorderColor': '#003D6B', 'secondaryColor': '#50E6FF', 'secondaryTextColor': '#323130', 'secondaryBorderColor': '#00BCF2', 'tertiaryColor': '#F2F2F2', 'tertiaryTextColor': '#323130', 'lineColor': '#0078D4', 'textColor': '#323130', 'mainBkg': '#DEECF9', 'nodeBorder': '#0078D4', 'clusterBkg': '#F2F2F2', 'clusterBorder': '#003D6B', 'titleColor': '#003D6B', 'edgeLabelBackground': '#FFFFFF', 'fontSize': '14px'}}}%%
flowchart TB
  classDef l1 fill:#DEECF9,stroke:#0078D4,color:#323130
  classDef l3 fill:#DFF6DD,stroke:#107C10,color:#323130,font-weight:bold
  classDef l4 fill:#FFF4CE,stroke:#FFB900,color:#323130
  classDef l45 fill:#E8D5F5,stroke:#8764B8,color:#323130
  classDef l48 fill:#E8D5F5,stroke:#8764B8,color:#323130
  classDef l49 fill:#FDE7E9,stroke:#D13438,color:#323130
  classDef l5 fill:#FFF4CE,stroke:#FFB900,color:#323130
  classDef l6 fill:#F2F2F2,stroke:#605E5C,color:#323130

  subgraph Row1["Foundation"]
    direction LR
    L1["L1 Tool Agent"]:::l1 -.-> L2["L2 Autonomous Agent"]:::l1 -.-> L3["L3 Self-Regulating ★"]:::l3
  end

  subgraph Row2["Advanced"]
    direction LR
    L4["L4 Adaptive General"]:::l4 -.-> L45["L4.5 Self-Architecting"]:::l45 -.-> L48["L4.8 Strategic Self-Model"]:::l48
  end

  subgraph Row3["Proto-AGI+"]
    direction LR
    L49["L4.9 Autonomous Strategic"]:::l49 -.-> L5["L5 Proto-AGI"]:::l5 -.-> L6["L6 Conscious Entity"]:::l6
  end

  Row1 --> Row2 --> Row3

Level Documents¶

Cumulative Safety Mechanisms by Level¶

%%{init: {'theme': 'base', 'themeVariables': {'primaryColor': '#0078D4', 'primaryTextColor': '#003D6B', 'primaryBorderColor': '#003D6B', 'secondaryColor': '#50E6FF', 'secondaryTextColor': '#323130', 'secondaryBorderColor': '#00BCF2', 'tertiaryColor': '#F2F2F2', 'tertiaryTextColor': '#323130', 'lineColor': '#0078D4', 'textColor': '#323130', 'mainBkg': '#DEECF9', 'nodeBorder': '#0078D4', 'clusterBkg': '#F2F2F2', 'clusterBorder': '#003D6B', 'titleColor': '#003D6B', 'edgeLabelBackground': '#FFFFFF', 'fontSize': '14px'}}}%%
flowchart TD
  classDef l1 fill:#DEECF9,stroke:#0078D4,color:#323130
  classDef l3 fill:#DFF6DD,stroke:#107C10,color:#323130
  classDef l4 fill:#FFF4CE,stroke:#FFB900,color:#323130
  classDef l45 fill:#E8D5F5,stroke:#8764B8,color:#323130
  classDef l48 fill:#E8D5F5,stroke:#8764B8,color:#323130
  classDef l49 fill:#FDE7E9,stroke:#D13438,color:#323130
  classDef l5 fill:#FFF4CE,stroke:#FFB900,color:#323130

  subgraph L1S["Level 1"]
    direction LR
    S1["Input validation · Error handling"]:::l1
  end

  subgraph L2S["Level 2 (adds)"]
    direction LR
    S2["Persistent world model · Goal priority"]:::l1
  end

  subgraph L3S["Level 3 (adds)"]
    direction LR
    S3A["Identity hash · Delta clamp 0.05"]:::l3
    S3B["Prediction gate · Escalation guard"]:::l3
    S3C["Ethical Kernel · Value lock"]:::l3
    S3D["Lyapunov C(t) · Budget + degrade"]:::l3
    S3E["Belief graph · Affect · Survival"]:::l3
  end

  subgraph L4S["Level 4 (adds)"]
    direction LR
    S4A["BGSS ≥ 0.7 · ShadowAgent"]:::l4
    S4B["7-step mod protocol · Atomicity"]:::l4
    S4C["Strategy suppression · Skill lifecycle"]:::l4
    S4D["Growth-stability zones · 6 meta-procs"]:::l4
  end

  subgraph L45S["Level 4.5 (adds)"]
    direction LR
    S5A["Existential Guard · Frame veto"]:::l45
    S5B["Graduated recomp · ROD depth 3"]:::l45
    S5C["SEOF ensemble · Purpose coherence"]:::l45
    S5D["Fragmentation detect · 8 invariants"]:::l45
  end

  subgraph L48S["Level 4.8 (adds)"]
    direction LR
    S6A["Probabilistic world model"]:::l48
    S6B["Causal reasoning · Capability matrix"]:::l48
    S6C["Multi-horizon planning · Budget opt"]:::l48
  end

  subgraph L49S["Level 4.9 (adds)"]
    direction LR
    S7A["Autonomous goal gen · Value sandbox"]:::l49
    S7B["Resource survival · Multi-agent belief"]:::l49
    S7C["Trust calibration · Autonomy verify"]:::l49
  end

  subgraph L5S["Level 5 (adds)"]
    direction LR
    S8A["Persistent identity 10K+"]:::l5
    S8B["Cross-domain · Goal ecology"]:::l5
    S8C["Existential resilience · Self-reconstruct"]:::l5
  end

  L1S -.-> L2S -.-> L3S -.-> L4S -.-> L45S -.-> L48S -.-> L49S -.-> L5S

Key Metrics Summary¶

Reading Guide¶

• New to MSCP? Start with the MSCP Overview for a conceptual overview, then read Level 1 → Level 3
• Interested in safety? Focus on Level 3 (sections 4, 6, 9) and Level 4.5 (Phase V: Existential Guard)
• Interested in self-improvement? Focus on Level 4 (sections 5–7) and Level 4.5 (Phases I–II)
• Interested in strategic planning? Focus on Level 4.8 (Phases 1–3) for world modeling and strategic layer
• Interested in autonomous agency? Focus on Level 4.9 (Phases 1–5) for autonomous goal generation and value evolution
• Interested in AGI architecture? Focus on Level 5 for persistent identity and cross-domain generalization
• Interested in affect/emotion? Focus on Level 3 (section 7) for the foundational design

Foundational References¶

Core academic works referenced across the MSCP Level Series:

Full reference lists are provided at the end of each level document.

See the Home for project overview and repository structure.