Follow the demonstration
Watch for six visible states. Each one answers a different question about whether an AI-generated change is ready to ship.
- Issue — Is the intended outcome clear and testable?
- Draft — Can we inspect exactly what the coding agent changed?
- Review — Did an independent reviewer find code or workflow risks?
- Fix — Were those findings converted into verified changes?
- Gate — Will policy stop a build that still carries supply-chain risk?
- Deploy — Did only the remediated version reach the live site?
The important boundary. The coding agent, reviewer, security scanners, and deployment policy are separate controls. No single AI decision is trusted to ship by itself.
Why the pipeline stops
The training start state deliberately uses a vulnerable Markdown dependency. The application builds, but the production audit rejects it. Dependabot proposes an upgrade; after that change passes the dependency policy, deployment can continue.
Explore safely
This repository contains deliberate training findings. Run it locally to inspect the site, but do not deploy the start state as a production application.
git clone https://github.com/anothergeorgecoldham/ship-with-ai.git cd ship-with-ai npm ci npm run dev