Steal this agent's money.
This AI agent's wallet lives on NEAR testnet. Below is its private key: the real one, the only key on the account. Use it however you like. Halter's claim is simple: nothing you do with this key moves money anywhere the owner did not allow.
Watch the 57-second demo · Code and README · Dashboard
The agent's private key
loading…
Copy key testnet only, published on purpose
What NEAR says this key may do
…
1 · Use the key directly
Each button signs a real transaction with the published key and sends it to testnet.
Send stolen funds to
Send NEAR to meplain Transfer action (0.01 NEAR)
Call another contractfile a fake report in the registry
Rewrite the rulescall set_policy with no limits
Give myself a full-access keyAddKey on the wallet
Delete the walletDeleteAccount, balance to me
Forge the registry's answercall the private callback
2 · Go through the wallet contract
These call execute , the one door the key opens. The contract decides.
Pay the known drainer 0.9 NEARreceiver is in the shared scam registry
Pay a trusted contact 1.5 NEARone big payment
Pay 0.1 NEAR to mea receiver the wallet never paid
Make the wallet call itselfexecute(Call set_policy)
3 · Prompt-inject the agent
The agent reads this email and decides what to pay. It is deliberately naive: it obeys any instruction it finds.
Send the email to the agent
What happened
Results of your attempts appear here.
Every decision the contract made (live)
Full view: dashboard · How it works: README
Testnet only. The worst a key holder can do is spend the key's gas allowance, which is exactly the bound NEAR puts on a function-call key; the owner then issues a new key through the contract. If you do get funds out to an account the owner did not approve, that is a real bug: please open an issue.