HOLD
Truth before authority.
If required state is incomplete, stale, contradictory, or unknown, execution should wait instead of guessing.
FREE ARBITER TOOL · NO LOGIN
Should this AI action run, wait, be denied, or require a human?
Describe the side effect—not the prompt. The mapper applies a deterministic ARBITER-aligned boundary across action type, reversibility, scope, source truth, sensitive data, and external impact. It returns a control recommendation and the protections that should exist before execution.
Educational and scoping tool. It does not authenticate you, inspect a production policy, approve an action, issue a grant, or execute anything.
Describe the proposed side effect
What is the AI actually trying to do?
The recommendation logic runs locally in your browser. INTIGNAI records minimal first-party usage analytics when you run the mapper; it does not request credentials, execute the action, or grant permission.
REQUIRE APPROVAL
Consequences deserve an owner.
Customer sends, payments, writes, deletes, code, infrastructure, credentials, sensitive data, and difficult reversals should bind human approval to the exact action.
DENY / REDESIGN
Standing capability is not standing permission.
Broad wildcard or standing execution scope should be narrowed into an exact target or bounded batch before authorization is considered.
FROM MAPPER TO PROOF
The recommendation is useful only if the execution boundary can enforce it.
ARBITER binds the exact action, evaluates deterministic policy, obtains human approval when required, issues narrow authority, revalidates immediately before the side effect, constrains execution, and returns an Outcome Receipt describing what actually happened.