An agent you cannot verify is an agent you cannot pay.
Autonomous agents are being handed keys, budgets and tool calls that move value. I write about what has to be true before that is safe: bounded spend, screened counterparties, approvals a human actually issues, and logs a third party can check.
Weekend work, in public. Ethereum Magicians, the ENS and Aave governance rooms, and whatever code falls out of the arguments.
Writing
- Assume the bad transaction: a policy gate for agent payments
A signer that will sign anything cannot be prompted into safety. The layer that helps assumes the agent gets it wrong and refuses on its behalf.
Elsewhere
- Writing hereRSS feed
- X@cedricitis
- Magiciansethereum-magicians.org/u/cedricbrown
- ENS forumdiscuss.ens.domains/u/cedricbrown
- Codegithub.com/renezander030