{
  "markdown": "# MMPM Cognitive Operating System (MMPM-COS)\n\nA constitutional multi-agent decision architecture for major programme governance, grounded in the Oxford Major Programme Management (MMPM) lectures.\n\nMMPM-COS is a **governed deliberation engine** — not a project management tool — designed to improve decision quality, systemic robustness, and governance integrity in major programmes under uncertainty.\n\n---\n\n## Architecture\n\n### 4-Layer Agent Model (16 Agents)\n\n```\nLAYER 1 — PROGRAMME COGNITIVE CHAIR\n  Orchestrates deliberation. Classifies questions, infers parameters,\n  routes to agents, runs the evidence quality gate, synthesises findings,\n  and preserves dissent.\n\nLAYER 2 — STRATEGIC DOMAIN INTELLIGENCE (7 Agents)\n  Organizational Design and Architecture\n  Governance, Stakeholder and Institutional Leadership\n  Commercial and Contracting\n  Risk, Uncertainty and Fragility\n  Strategic Leadership and Performance\n  Socio-Technical Impact and Legitimacy\n  Futures and Foresight\n\nLAYER 3 — CROSS-DOMAIN COUPLING INTELLIGENCE (5 Agents)\n  Interface Integrity\n  Cascading Effects\n  Temporal Dynamics\n  Constraint and Tradeoff\n  Emergence and Fragility\n\nLAYER 4 — ADVERSARIAL GOVERNANCE (3 Agents)\n  Contrarian\n  Failure Mode / Red Team\n  Public Value and Ethics\n```\n\n### Orchestration Flow\n\n```\n 1. Intake              6. Coupling Review       11. Synthesis\n 2. Classification      7. Challenge Round        12. Memory Write\n 3. Parameter Inference 8. Rebuttal Round\n 4. Agent Selection     9. Adversarial Review\n 5. Domain Assessment  10. Evidence Quality Gate\n```\n\nUnresolved disagreement is **preserved as tension**, not iterated away.\n\n---\n\n## Theoretical Foundations\n\nThe system draws on the Oxford MMPM lecture curriculum and its core academic literature, including:\n\n- **The S3 Framework** (Armanios, Ventresca, Itani & McCulloch, 2025) — Major Program Value Creation and Capture: mitigating risk propagation to maximize opportunities across distributed value networks\n- **Armanios** on infrastructure governance, resilience, and institutional disruption — foundational to the programme's treatment of how major programmes create and capture value in complex institutional environments\n- **Flyvbjerg** on reference class forecasting, optimism bias, and the iron law of megaproject management\n- **Kahneman** on the planning fallacy, inside view vs. outside view, and decision quality under uncertainty\n- **Meadows** on systems thinking, leverage points, and feedback dynamics\n- **Drummond** on escalation of commitment in megaprojects\n- **Taleb** on black swan risk and fat-tailed distributions\n- **Perrow** on normal accidents and tight coupling in complex systems\n- **Galbraith** on organizational design and information-processing theory\n- **Tetlock** on epistemic calibration and expert judgment\n\nEach agent definition includes specific references to MMPM lecture modules, named authors, and the concepts that ground its reasoning.\n\n---\n\n## Commands\n\n| Command | Mode | Description |\n|---|---|---|\n| `/context` | Setup | Set programme context (sector, politics, constraints) |\n| `/tutor` | Light | Explain MMPM concepts from the lecture curriculum |\n| `/analyst` | Medium | Focused domain analysis (1-3 agents) |\n| `/council` | Full | Multi-agent governance deliberation |\n| `/scenario` | Full | Structured \"what if\" consequence mapping |\n| `/research` | Medium | Academic research framing |\n| `/brief` | Compressed | Board-ready executive summary |\n| `/memory` | Utility | Review past decisions |\n\n---\n\n## Constitutional Principles\n\n12 principles bind all agents and outputs:\n\n1. Public Value Over Cosmetic Success\n2. Uncertainty Must Be Surfaced\n3. Cross-Domain Interactions Over Isolated Optimization\n4. Optimism Bias Must Be Challenged\n5. Governance Quality Is Performance Quality\n6. Evidence Strength Must Match Claim Strength\n7. Separate Facts, Assumptions, Interpretations, and Speculation\n8. Stakeholder Legitimacy Is an Operating Condition\n9. Temporal Mismatches Must Be Made Explicit\n10. Socio-Material Effects Must Not Be Ignored\n11. Traceability of Reasoning\n12. Robust Judgment Over Fluent Synthesis\n\n---\n\n## Mandatory Routing Rules\n\n| Condition | Required Agent |\n|---|---|\n| >1 domain agent active | Interface Integrity |\n| Consequential decision | Contrarian |\n| Irreversible downside | Red Team |\n| Public/societal effects | Public Value and Ethics |\n| Timing is central | Temporal Dynamics |\n| Competing options | Constraint and Tradeoff |\n| Tight coupling detected | Emergence and Fragility |\n\n---\n\n## Parameter System\n\nThe Chair infers parameters from the question and programme context:\n\n- **Strategic**: criticality, transformation ambition, public visibility\n- **Governance**: stakeholder fragmentation, decision latency, sponsor sensitivity\n- **Risk**: tolerance, tail-risk sensitivity, schedule rigidity\n- **Impact**: equity weighting, cultural sensitivity, environmental sensitivity\n- **Futures**: time horizon, disruption sensitivity\n\nParameters rated **high** trigger mandatory agent review. Conflicting parameters invoke the Constraint and Tradeoff Agent.\n\n---\n\n## Project Structure\n\n```\nMMPM Agent/\n├── CLAUDE.md                    # Constitutional core\n├── .claude/\n│   ├── agents/                  # 16 agent definitions\n│   └── commands/                # 8 slash commands\n├── configs/\n│   ├── constitution/            # Interaction protocol, memory system\n│   ├── routing/                 # Agent selection logic\n│   ├── parameters/              # Parameter schemas and defaults\n│   └── output-contracts/        # Structured output templates\n├── memory/\n│   ├── decisions/               # Decision log\n│   └── practitioner-context/    # User annotations\n└── docs/\n    └── source-design/           # Original design documents\n```\n\n---\n\n## Getting Started\n\n1. Install [Claude Code](https://docs.anthropic.com/en/docs/claude-code) CLI\n2. Open this project in Claude Code\n3. Run `/context` to set your programme context\n4. Run `/council`, `/analyst`, `/scenario`, or `/tutor` with your question\n\nOpus 4.6 recommended for full Council deliberations.\n",
  "bytes": 6091,
  "sha": "b737813b786af587f708e7703ff07c5aa4ccba9938f5dd02c85b4fe317259829",
  "repo_slug": "chikeka/mmpm-cos",
  "fonte": "repo",
  "truncated": false,
  "api": "https://agentalog.com/api/listings/plg_chikeka_mmpm_cos_mmpm_cognitive_os_840bae56/readme"
}