NODE MS-R01 DOMAIN SOVEREIGN AI CONTROL MODEL LEAST AGENCY STATE ACTIVE

PUBLIC EVIDENCE VAULT / VERIFIED ARTIFACTS

PROOF

Claims are cheap. This is the public evidence trail: released systems, working code, and published research.

SOURCE INDEX / github.com/themayursinha

CODE Runtime tools and labs CONTEXT Architecture notes RESEARCH Published work MODE Inspectable

01 / RELEASED SYSTEMS

CONTROL INFRASTRUCTURE

2026 Builder PUBLIC ARTIFACT

MCP Visor

Deterministic, non-AI policy engine that intercepts MCP tools/call before execution. Runtime policy enforcement and audit control plane for tool-using AI agents.

GitHub Context Paper

2026 Builder PUBLIC ARTIFACT

Agent Identity Plane

Cryptographic identity and actor-chain provenance for AI agents. A self-hosted agent registry, token service, and verifier that issue short-lived single-audience tokens carrying the full user-to-agent-to-tool delegation chain, so downstream systems can answer who acted, for whom, and through which agents. v1.0.

GitHub

2026 Builder PUBLIC ARTIFACT

Authority Graph Simulator

Counterfactual authorization for agent actions. Evaluates a proposed capability delegation against a declared authority graph and denies it when the authority an agent could reach would exceed the originating mandate.

GitHub

2026 Builder PUBLIC ARTIFACT

Capability Delta Receipts

Hash-linked receipts that record whether each step of an agent trajectory gave it a materially stronger primitive, and pause before the first effect outside the declared envelope.

GitHub

2026 Builder PUBLIC ARTIFACT

Vulnerability Research Harness

Deterministic campaign harness for authorized AI-assisted vulnerability research. Pins the target and success condition in a campaign contract, prevents premature agent convergence with an approach registry, and records every hypothesis-to-impact transition in an evidence ledger.

GitHub

2026 Builder PUBLIC ARTIFACT

CTEM Leader Lab

Interactive implementation workbench for security leaders moving from vulnerability management to Continuous Threat Exposure Management.

GitHub Demo

2025 Builder PUBLIC ARTIFACT

MCP security evaluator

Companion tooling for evaluating MCP server trust boundaries, tool metadata drift, and LLM integration risks.

GitHub Context

2025 Builder PUBLIC ARTIFACT

Adversarial ML security lab

Hands-on lab for adversarial ML attack and defense experiments.

GitHub

02 / PUBLISHED RESEARCH

EVIDENCE BASE

View full publication list on ResearchGate →

  1. MCP Visor: Deterministic Runtime Enforcement for Governing Tool-Using AI Agents
  2. LLM NL2SQL for HR analytics
  3. BERT-based RASP for runtime security
  4. Fine-tuning BERT for HTTP payload classification
  5. Predictive Kinetic Threat Modeling via Deep Neural Networks
  6. Asymmetric Threat Mitigation via Autonomous UAVs
  7. A game-theoretic approach to evaluate security controls
  8. Entropy Aware Feedback Directed Fuzzing for Vulnerability Discovery
  9. Understanding IoT botnets