This repository intentionally keeps two versions of the prototype side by side.
The first version demonstrates the happy-path idea:
Prompt -> Intent -> Spec -> Mock Payload
It proves the basic concept of converting natural language into a structured infrastructure automation artifact.
The second version introduces a hard safety boundary before the Spec Agent:
Prompt -> Safety Gate -> Intent -> Spec -> Mock Payload
The safety gate blocks unsafe inputs before downstream automation logic runs.
The most important design decision is not the Python implementation.
The most important design decision is the boundary:
Unsafe natural-language intent should not become a buildable automation plan.
That is the difference between a demo and an enterprise delivery architecture.
In Professional Services, customers often begin with ambiguous requests.
A strong delivery engineer must translate that ambiguity into:
- Validated inputs
- Clear assumptions
- Known risks
- Defined rollback behavior
- Reviewable designs
- Evidence-ready outputs
This prototype demonstrates that mindset in code.