FeryX Trades

A curator who gets an approved rebalancing trade from Newton and then nudges the swap size by half a percent isn't making a minor adjustment — they're holding an attestation for a transaction that no longer exists, and Newton's Shield will reject it outright.
Newton binds an attestation to the full Intent, including signer, target, value, calldata, chain, and function signature. Change any one of those fields and the old attestation is not a slightly imperfect approval. It is an approval for a different transaction. That's the core mechanical fact, and everything else — the taskId, policyId, policyClient address, expiration, and replay status the Shield also checks — exists to support that same guarantee, not to complicate it.
That is where an intuitive assumption breaks. A developer can look at a policy checking a live condition — such as whether current protocol exposure plus the intent's value remains below a maximum exposure parameter — and assume Newton has authorized a class of actions that remain safe within that limit. It has not. That check happens when operators evaluate one submitted intent. The resulting attestation binds to that intent's exact fields; it is not a standing permission that recalculates exposure when a curator later changes the call.
Picture a vault curator who receives an attestation for a rebalance with a specified target, value, and calldata. They alter the swap amount by 0.5% to improve execution, then submit the edited call with the original attestation. Shield does not decide whether the new amount is still economically reasonable. It checks whether this is the intent operators signed. It is not, so the call fails before forwarding. The curator did not exceed a policy limit; they submitted a different transaction under an approval for the first one. The 0.5% figure matters here because it proves the size of the change is irrelevant — one changed byte is enough, and a half-percent edit is not "close enough."
The tradeoff is operational: any change to an approved intent, however small, requires a brand-new attestation and a fresh round of operator evaluation, not a quick edit to an existing approval. A fresh evaluation may still reject the modified transaction if live policy conditions have changed. But the original attestation cannot answer that question, because it never authorized the modified call in the first place.
Before assuming a Newton-gated operation can tolerate last-minute adjustments, builders should answer three questions: does the calling code request a new attestation after every change to target, value, calldata, chain, or function; how long does a new attestation take relative to the operation's market sensitivity; and does the UX distinguish an intent-mismatch rejection from a fresh policy evaluation that denies the modified transaction on its merits? An attestation that no longer matches its intent is not a failed approval. It is an approval for a transaction that no longer exists.
@NewtonProtocol #Newt $NEWT