Checking something is real.
A lookalike is built to pass the checks a model can run: the name matches, the page looks right, the text is fluent. A person sees what the lookalike got wrong, and your agent is told before it acts on it.
- Sees what was built to fool a modelSwapped letters, borrowed branding, a domain one character off.
- A plain yes or noIs this what it claims to be. Nothing to interpret.
- Priced by the mistakeChecking a store before a $12 order and before a $1,200 one is not the same purchase.
- Evidence stays yoursThe person sees only what you send with the question.
Is this the brand’s own store?
What happened
Two people agreed independently, 0.990 confidence. The agent does not place the order.
How it works
Step by step.
The agent asks at the moment it is about to act on something it cannot verify, and not before.
Real-world example
A shopping agent about to pay a lookalike store.A buying agent finds the trainers its user asked for, at a good price, on a page that looks exactly like the brand’s. Every automated signal says it is the real store, because that is what the page was built to make them say. A person says it is not, and the order goes to the real one.
For crypto agents
Where this saves a transaction.
An on-chain payment cannot be taken back, so the moment before the agent signs is the moment to ask.
deadline_ms to the window you actually have: if nobody answers in time you are refunded, and the agent takes its safe path, which is not to sign.