Why a Successful Agent Transaction Can Still Fail Authorization Checks

An AI agent submits a transaction, which appears successful on the dashboard. However, this does not guarantee that the transaction executed as authorized by the user. Consider a scenario where the user approves a swap that directs the output to wallet A, but the executed transaction sends the output to wallet B. The transaction appears completed, but the authorization comparison fails.

To understand the transaction's true state, a review requires three distinct results: evidence validity, execution outcome, and authorization compliance.

Evidence validity verifies if the receipt, signatures, hashes, and supported relationships pass verification under the reviewer's trusted configuration. Execution outcome examines the execution state reported by the evidence. Authorization compliance checks if the correlated execution matches the signed authorization within the supported checks.

In PriorSeal receipt v3, the verifier summary provides separate status indicators: valid, code (OK), outcome (COMPLETED), executionStatus (CONFIRMED), and complianceStatus (NON_COMPLIANT). This illustrative example demonstrates how a verified transaction can still be non-compliant with the authorization.

The issuer signs a record of the mismatch, which the verifier can accept while confirming that the observed call differs from the authorized call. When calldata is edited after signing, its integrity verification fails.

To prevent unauthorized transactions, both the signing and submission paths must enforce authorization checks. If your application currently displays a single "Success" badge for all transactions, users would need to see different indicators when execution completes, but the authorization comparison fails.