Deterministic Pre-Execution Enforcement Layer for Web-Based AI Workflows (Reference Implementation)

Dear W3C team,

I am writing to share a working demonstration of a deterministic
pre-execution enforcement layer for AI systems. A presentation deck
containing two live demonstrations is attached.

The demonstrations illustrate constraint validation prior to execution
across (i) lending decision systems and (ii) multi-step AI workflows. This
approach implements a fail-closed validation layer, enabling deterministic
and auditable enforcement of predefined constraints before system actions
are executed.

As AI agents increasingly operate across web interfaces and APIs,
deterministic enforcement of permissible actions may become relevant to web
architecture and security considerations.

The demonstrations are live and publicly accessible, and the underlying
protocol specification is linked within the deck.

Thank you for your consideration.

Kind regards,
Ana Margarida Esteves