Your choices become characteristics
Each option maps to entries in the published characteristic taxonomy — things like "handles sensitive data" or "takes an action that cannot be undone".
For a first project
Don't know how to describe it? Just pick.
A few taps instead of a paragraph. Choose what you are building and AgentDirective shows the safeguards that apply, using the same published rules as the main review. These choices are repeatable, and you can review them before continuing.
Already know what you want to build? Describe it in your own words instead — a description can capture details a menu may miss.
Step 1
What are you building?
Pick the closest match. You can correct everything on the next screen.
Your starting point
What applies to this
What this is, and what it is not
Next you confirm these characteristics, then answer the decisions only you can make.
Why this is reproducible
How a tap becomes a safeguard
Every option carries a fixed set of architecture characteristics. No model is involved at any point.
Rules decide what applies
The same published rules used in the main review decide which Control Packs apply to those characteristics.
You confirm before it counts
On the next screen, you can keep or remove each characteristic. Choosing an option does not make a decision for you.
The decisions stay yours
Amounts, scopes, and approval boundaries remain open until you confirm them. The guided path leaves those decisions for you.