Engineering Governance
AGD Methodology FAQ
Answers regarding Assurance-Gated Development, bounded AI agent delegation, verification protocols, and release governance.
What is Assurance-Gated Development (AGD)?
Assurance-Gated Development (AGD) is an engineering methodology designed for high-assurance software engineering and supervised AI agent development. It structures engineering tasks into strictly bounded packages governed by three core gates: Specify, Authorize, and Verify.
Does AGD act as a security sandbox or automated isolation tool?
No. AGD is an explicit development methodology and validation process, not an operating system sandbox or runtime security jail. It enforces rigorous boundary definitions, verifiable evidence logging, independent test execution, and non-delegable human release authority before any code reaches production.
Where can I download the AGD Community Edition?
The AGD Community Edition reference implementation is undergoing independent review, with release and licensing decisions still pending. No public download, licensing terms, or release date is currently announced. Links will be added only if publication is expressly authorized.
Why is human release authority a mandatory principle in AGD?
While automated tools and AI coding agents can execute bounded tasks and run test suites, final release decisions require human accountability. Under AGD, no autonomous agent is permitted to deploy to production, publish packages, or modify live infrastructure without explicit human review and authorization.