Exposure Map

Where can your agents act without a floor?

Four questions about your agent architecture. No email, nothing stored, instant result — an indicative coverage map of where an agent can take an irreversible action without a deterministic floor. Not a certification.

1 / 4 · Write reach (least privilege)

Can your agents natively write, send, or spend — or are write and destructive tools withheld until explicitly granted?

2 / 4 · Fail-closed defaults

On error, timeout, or low confidence, does the agent fail closed (stops and escalates) or fail open (proceeds or guesses)?

3 / 4 · Human commit gate

Is the human gate a soft “approve” button — a rubber stamp — or a schema-locked commit the agent cannot satisfy on its own?

4 / 4 · Identity & append-only audit

When an agent executes an action, how is the transaction recorded?

Answer all four to see your indicative exposure map.