Your AI Agent Says the Work Is Done. Verify It in Your Own Stack.

Description and use

Implementing the AER‑1 receipt format enables verification that AI tool outputs remain unchanged after creation, providing transparency and auditability.

Problem

When an AI agent reports success, there is no reliable way to confirm that the saved result is the same as the one originally produced, leading to potential silent rewrites or edits.

Outcome

Implementing AER‑1 receipts provides a lightweight, local method to confirm that AI tool outputs have not been altered after creation, reducing audit risk and enabling independent verification.

Workflow

• Fetch the conformance corpus from https://zambo.dev/aer1/corpus.json
• Implement canonicalization: sort keys, remove whitespace, serialize values in shortest stable form, hash with SHA‑256
• Verify each record’s hash matches the stored hash
• Recompute the chain by checking each record’s prev_hash against the previous record’s hash
• Run all 129 checks from the corpus and fix any failures
• Optionally use the reference tooling (pip install aer1-verify or npm install aer-1) to automate verification

Evidence

A receipt proves the saved result was not changed after the fact. It never claims the answer was right.

Source

The conformance corpus lives at https://zambo.dev/aer1/corpus.json. It carries 126 test vectors covering 129 checks, at kit version 1.5.7.

Source

pip install aer1-verify (version 0.2.20) or npm install aer-1 (version 0.1.25)

Source