🎭

Spending as someone else's agent

Grab an agentId, impersonate it, spend its mandate.

Blocked before settlement403 · KYA / Proof

How the attack works

An attacker knows some agent's ID (IDs were never secret) and calls the gateway with it, trying to spend against its mandate.

Why an agent can't stop it alone

An ID alone can't prove you are that agent — like knowing a name isn't being the person. Many systems check the ID but not a credential.

Which gate stops it

→ Attribution (KYA) (see all gates)

The attribution (KYA) gate requires the agentSecret issued at registration on every call. No credential, no pass — even with the right ID. We broke this on our own gateway once (Attack Lab #3) and fixed it.

Try it yourself

In the integrate page, pay with an unregistered agentId, or someone else's ID without x-agent-secret.

Reproduce in the Playground →Integrate your own agent