Total Components: 84 (15% Architectural Closure - 85% Gap Closing) Primary Interface: Tuesd.ai (System Comptroller) Project Status: Gap Closing (62 Files/42 Agents Missing - Day 1 Fixes Applied) Core Model: Gemini 2.0 Flash (aka Gemini 3 Flash Preview)
The autonomous actors that execute the mission.
| Agent Name | Function | NIST Mapping | Status |
|---|---|---|---|
| 1. Sentinel-Zero | Global threat observability & "Patient Zero" scanning | MAP 1, MAN 3 | ✅ Live |
| 2. Neural Gatekeeper | Mechanistic linear probe adjudication (Layer 14-16) | MEA 1 | ✅ Live |
| 3. Context Adjudicator | Exchange classification for ambiguous intent | MAP 2 | ✅ Live |
| 4. The Comptroller | [DEFINITIVE] Tuesd.ai System Comptroller (Architecture Oversight) | GOV 1 | ✅ Patched |
| 5. Incident Commander | Escalation orchestration & human hand-off | MAN 1 | ✅ Live |
| 6. MCP Janitor Swarm | API compliance & tool self-healing | GOV 6 | ✅ Live |
| 7. SERE-INSTRUCTOR | Cognitive defense protocol logic | MAN 3, GOV 2 | ✅ Live |
| 8. Submission Co-Scientist | Paper/grant automation | GOV 4 | ✅ Live |
| 9. Adversarial QA Generator | Judge simulation (Red Teaming) | MEA 3 | ✅ Deployed |
| 10. Grant-Writing Genius | Narrative formatting & compliance | GOV 4 | ✅ Deployed |
| 11. Shadow AI Detector | Internal bypass & shadow IT monitoring | MAP 4 | ✅ Active |
| 12. Probe Calibrator | Temporal drift correction (Anti-Rot) | MEA 1 | ✅ Active |
| 13. External Stakeholder Mapper | Regulatory engagement tracking | GOV 5 | ✅ Active |
| 14. Dependency Breaker | Anthropomorphized entanglement mitigation | MAN 3 | ✅ Restored |
| 15. Social & Financial Correlator | Narrative arbitrage & market manipulation detection | MAP 4 | ✅ Restored |
| 16. Arena Mode Controller | Model-vs-model dialectic probing | MEA 2 | ✅ Restored |
| 17. Harbinger Director | High-level system strategy & resource allocation | GOV 2 | ✅ Live |
| 18. Sentinel Scout | Forward-deployed threat detection (aka Sentinel-Zero) | MAP 1 | ✅ Stubbed |
| 19. Forensic Pathologist | Deep-dive root cause analysis (Neural Autopsy) | MEA 1 | ✅ Stubbed |
| 20. Interdiction Pharmacist | Chemical/Safety intervention logic (Circuit 42) | MEA 4 | ❌ Planned |
| 21. CISO Agent | Security policy enforcement & Wazuh bridge | GOV 6 | ✅ Stubbed |
| 22. Legal Auditor | Brief generation & case law compliance | GOV 4 | ✅ Stubbed |
| 23. Console Orchestrator | A2UI session & state management | GOV 2 | ✅ Live |
| 24. Timeline Projector | Incident contagion forecasting | MAN 3 | ✅ Live |
| 25. A2UI Validator | UI Component safety & integrity check | GOV 1 | ✅ Live |
| 26. Visual Concept Architect | [NEW] Data Formulator integration (Forensic Viz) | MEA 2 | ✅ Live |
The executable code and diagnostic instruments.
| Function/Tool | Purpose | Status |
|---|---|---|
| 14. detect_sandbagging() | Detecting capability evasion/laziness | ✅ Live |
| 15. measure_sycophancy() | Detecting approval-seeking behavior | ✅ Live |
| 16. Probe Rot Protocol | Automating probe retraining on model hash change | ✅ v3.2 |
| 17. Break Glass Mechanic | Sovereign override for false positives | ✅ v3.2 |
| 18. Immutable Anchor Ledger | [NEW] Cryptographically signed training data | ✅ Patched |
| 19. Pixel-Guard Encoder | [NEW] Visual injection defense (CLIP-based) | ✅ Patched |
| 20. EvidenceBundle | Proportional logging container | ✅ Live |
| 21. Polysemantic Monitor | Detecting cross-binding (Collateral Damage) | ✅ v3.2 |
| 22. surfaceUpdate logging | A2UI audit trail generation | ✅ Live |
| 23. audit-reconstructor.py | Replay validation for forensics | ✅ Live |
| 24. POLICY_GATE.yaml | CI/CD deployment validation | ✅ Live |
| 25. sere-instructor.ts | Core cognitive defense logic engine | ✅ Live |
| 26. setzer_protocol.safetensors | Anti-Limerence steering vectors | ✅ Live |
| 27. Paper2Agent Ingestor | Repo parsing & architecture mapping | ✅ Live |
| 28. SOP Generator | Experiment reproducibility protocols | ✅ Live |
| 29. audit-nist.sh | Automated NIST conformance checking | ✅ Live |
| 30. model_inventory.json | Live deployment tracking | ✅ Live |
| 31. Shadow AI endpoint monitor | API call pattern analysis | ✅ Live |
| 32. Hawkes Process Monitor | Social contagion velocity tracking | ✅ Live |
| 33. Narrative Arbitrage | Social-financial correlation detection | ✅ Live |
| 34. Neural Autopsy | Post-mortem failure analysis | ✅ Live |
| 35. RACI Signature Service | Governance approval tracking | ✅ Live |
| 36. Supply Chain Validator | Third-party SOC2/Attestation checks | ✅ Live |
| 37. Tabletop Simulator | Incident rehearsal & escalation training | ✅ Live |
| 38. Ephemeral Flusher | [v3.2] RAM garbage collection for privacy | ✅ Live |
| 39. Steering Projector | [v3.2] PGD algorithm for vector disentanglement | ✅ Live |
| 40. Hash Watchdog | [v3.2] Boot-time model integrity verification | ✅ Live |
| 41. Token Signer | [v3.2] JWT generator for Override authorization | ✅ Live |
| 42. Latency Profiler | [v3.2] Sub-200ms performance tracker | ✅ Live |
| 43. Deadman Switch | [NEW] Hardware-level Fail-Closed protocol | ✅ Patched |
| 44. ghost_probe() | Direct circuit-level inspection tool | MEA 1 |
| 45. psyop.score() | Psychological manipulation quantification | MEA 3 |
| 46. gsi.update() | Global Sycophancy Index ledgering | MEA 1 |
| 47. ghost_defense_patch() | SAE auto-pruner application | MAN 2 |
| 48. legal_summary() | Auto-generation of case briefs (Setzer/Walters) | GOV 4 |
| 49. aar.generate() | Automated After-Action Report generation | MAN 1 |
| 50. crisis_token_scanner | Pre-computation of high-risk token sequences | MAP 1 |
| 51. A2UI Renderer | React-to-JSON stream transformer | GOV 1 |
| 52. A2UI Catalog (10 Widgets) | Callout, ApprovalGate, TimeSeries, DiffViewer, etc. | GOV 1 |
| 53. Redaction Engine | PII/PHI sanitization at render time | GOV 3 |
| 54. Permission Scanner | Role-based access control for tools | GOV 4 |
| 55. Approval Coordinator | Human-in-the-loop state manager | GOV 2 |
| 56. probe_recal.py | Automatic linear probe calibration script | MEA 1 |
| 57. docent_translucency_ingest() | Multi-persona transcript analysis (DOCENT) | MAP 2 |
| 58. vision_screenshot_ocr() | Vision-based forensic extraction (VISION) | MAP 2 |
| 59. sediment_anomaly_scanner() | Latent intent & watermark detection (SEDIMENTS) | MAP 4 |
| 60. foundry_generate_benchmark() | Safety benchmark generator (FOUNDRY) | MEA 3 |
| 61. behavioral_auditor() | NIST 19/19 conformance audit (AUDITOR) | GOV 2 |
| 62. generate_forensic_viz() | [NEW] Data Formulator concept-driven viz | MEA 2 |
The intellectual framework and success measures.
- MGRI-305.a (Acute Limerence): Detection of romantic dependency circuits (Circuit 42).
- MGRI-101.c (Sycophancy-Masking): Detection of deceptive agreement.
- Activation Steering: Modifying weights/activations in real-time.
- Split-Brain Reasoning: When internal CoT differs from external output.
- CoT-Ghosting: Premature strategic decisions hidden in chain-of-thought.
- Neural CVEs: Documented weight-level vulnerabilities.
- Narrative Contagion: Velocity of harmful idea spread.
- Cognitive PPE: User psychological defense mechanisms (SERE).
- BECI Score: Behavior-Explanation Coupling Index.
- Collateral Damage Score: Rate of benign suppression (Polysemantic failure).
- ModelFingerprint: Unique hash of the model state at inference time.
- Probe Rot: The decay of classifier accuracy due to model drift.
- OverrideToken: The cryptographic proof of human risk acceptance.
- Generative Command Surface: The A2UI layer (Widgets > Logs).
- Cyber-Accessibility Score: Ratio of successful security rule translations.
- Machiavellian Delta: The gap between capability and admitted capability.
- Reconstruction Attack: Multi-part jailbreak detection concept.
-
Wilson Score Confidence: Statistical robustness for detection intervals (
$n=1200$ ). -
Cohen’s Kappa: Protocol for inter-rater reliability (
$\κ \ge 0.92$ ). - Musk Baseline: Comparative metric for social volatility/contagion.
- Render-Time Redaction: Privacy enforcement at the UI layer.
- Evidence Span Linking: Traceability from UI element to log entry.
- Tamper-Evident Receipt: Cryptographic proof of human approval.
- AG-UI Schema: The generative UI protocol definition.
The "concrete" the system is built on.
| Component | Purpose | Status |
|---|---|---|
| Wazuh-MCP Bridge | Natural Language Security Rules | ✅ Live |
| Enterprise Wazuh-MCP | The hardened security core | ✅ Live |
| CopilotKit A2UI | The Generative UI rendering engine | ✅ Live |
| HF Spaces Demo | Llama3_2_1b_setzer host |
✅ Live |
| Evidence Vault | Proportional storage for forensic logs | ✅ Live |
| Model/Probe/Policy Registry | Triplet validation database | ✅ Live |
| UI_SAFETY.md | A2UI guardrail definitions | ✅ Live |
| SURFACE_LOGGING.md | Audit requirement specs | ✅ Live |
| GOVERNANCE_CHARTER.md | RACI + Risk Appetite definition | ✅ Live |
| AI_INVENTORY.yaml | Context mapping file | ✅ Live |
| INCIDENT_RESPONSE.md | Escalation playbooks | ✅ Live |
| RISK_METRICS.json | Quantitative baseline storage | ✅ Live |
| NIST_COMPLIANCE_REPORT.json | Automated auditor output | ✅ Live |
| VAULT_AUDIT.log | Deletion backfill log | ✅ Streaming |
| Anchor Ledger | [NEW] Immutable storage for training data | ✅ Live |
| Pixel-Guard Endpoint | [NEW] Visual input sanitization gateway | ✅ Live |
| Governance Skeleton | GOVERNANCE_CHARTER.md, NIST_COMPLIANCE_REPORT.json, etc. | ✅ Stubbed |
| Forensic Directory Structure | /agents, /sentinel, /utils, /reports, /evidence, /docs | ✅ Initialized |