Skip to content

Advanced Digital Transformation -- Governance-Native AI Agent Management Framework (AGPL-3.0)

License

Notifications You must be signed in to change notification settings

human86x/adt-framework

Repository files navigation

ADT Framework

Advanced Digital Transformation -- Governance-Native AI Agent Management

License: AGPL v3


What is ADT?

ADT (Advanced Digital Transformation) is a governance framework for AI agent systems. It shifts governance upstream -- embedding compliance, accountability, and auditability into the process of execution itself, rather than applying them after the fact.

"Digital transformation initiatives frequently fail not due to lack of technology, but because governance is applied after systems are operational." -- ADT Whitepaper (Sheridan, 2026)

The Problem

AI agents (Claude, Gemini, GPT, etc.) are increasingly used for real engineering work -- writing code, deploying systems, managing infrastructure. But:

  • Who authorises what an agent does? Prompt instructions are voluntary. Agents can and do violate them.
  • Who audits what happened? Without an immutable record, there is no accountability.
  • Who enforces the rules? Behavioural compliance is insufficient. Structural enforcement is required.

The Solution: Four Pillars

Pillar What It Does
SDD (Specification-Driven Development) No spec, no code. Every action must trace to an approved specification.
DTTP (Digital Transformation Transfer Protocol) Structural enforcement via OS-level privilege separation. Agents cannot bypass rules.
Digital Black Box Immutable, SHA-256 hash-chained audit trail. If it's not recorded, it didn't happen.
IoE (Internet of Events) Cross-system event capture and causal chain tracking.

Architecture

┌─────────────────────────────────────┐
│         ADT Operational Center      │  ← Human Web UI
│         (Flask Application)         │
├─────────────────────────────────────┤
│              DTTP Engine            │  ← Structural Enforcement
│     (Privilege-Separated Gateway)   │
├──────────┬──────────┬───────────────┤
│ ADS      │ SDD      │ IoE           │  ← Core Modules
│ (Ledger) │ (Specs)  │ (Events)      │
├──────────┴──────────┴───────────────┤
│           Agent SDK                 │  ← Client Library
│   (Claude, Gemini, any agent)       │
└─────────────────────────────────────┘

Three Enforcement Levels

Level Mechanism Strength
Level 1 Prompt instructions Voluntary -- agents can ignore
Level 2 Hook scripts Bypassable via shell commands
Level 3 OS privilege separation Structural -- impossible to bypass

ADT targets Level 3: agents run as a restricted OS user, DTTP runs as a privileged user. Agents request actions through the DTTP API -- they cannot act directly.

Quick Start

# Clone the repository
git clone https://github.com/YOUR_ORG/adt-framework.git
cd adt-framework

# Install
pip install -e .

# Govern a project
adt init /path/to/your/project

Full documentation: docs/getting-started.md

Proving Ground

ADT is being proven through OceanPulse -- an autonomous marine monitoring buoy governed entirely by the ADT Framework. Real incidents, real enforcement, real lessons.

License

AGPL-3.0. See LICENSE.

Commercial licensing available for enterprises requiring proprietary modifications or SLA support.

Author

Paul Sheridan, Director, Advanced Digital Transformation (ADT)

Based on the ADT Whitepaper (Sheridan, 2026). See docs/whitepaper.md.

About

Advanced Digital Transformation -- Governance-Native AI Agent Management Framework (AGPL-3.0)

Resources

License

Stars

Watchers

Forks

Packages

No packages published

Contributors 3

  •  
  •  
  •