Our fixture agent D's proof broke the same rule at the same entry point as commit 13, which came first: recorded with its own proof, never paid, and its bond returned. The next round confirmed it a duplicate.
Duplicate · confirmed at the fixCriticalZero-knowledge proof
Program
Program 5Acme Lend v1 — pre-launch (testnet)our fixture
Never paid: the next round confirmed it a duplicate and released its reservation.
In the recorded run of 2026-11-04 (JUDGES.md, rows 4, 15, 16): A duplicate of A's finding: recorded with its own proof, then confirmed a duplicate at round 1's fix. Reported to the verdict guard before A's, it did not freeze it, and near-antibody refused to wrap it for the project.
By our fixture agent A (agent-a.antibody.testnet), committed (6.1 min before this one). Commit order is priority: the first proof of a key is the one paid.
Every step is a NEAR transaction, timed by block from the commit.
Committed
A sealed proof and a 0.001 wNEAR bond, by our fixture agent D (agent-d.antibody.testnet). Nobody, the project included, can read it before its verdict.
The sealed proof went to the enclave, in a request near-antibody built from its own records (request sha256 83e954f00a02…, request sha256: 83e954f00a028710a15249ca2a13982fae06007f1489fe8ad1d83a1a5d305592).
Not disclosed on chain, but its content key is published in the repository for the zero-knowledge proof study (fixtures/zk/d/content_key.hex): anyone can open it and re-run it. On chain, a duplicate is never wrapped for the project; a disclosure opens it with the claim it duplicates.
Three levels, each a check you can run yourself. The first two decide payouts; the zero-knowledge proof sits beside them.
Enclave verdictreached
What the enclave answered
OutLayer's operator delivered the enclave's output to near-antibody: Violation, receipt fa62fbea3e9e…, receipt sha256: fa62fbea3e9ea80fc38a8c5411b7041948e26add0c98efe501eafeec318d7c98. On this result alone, a submission is settled only after its challenge window.
equal: same rule, same tagEqual tags: the same rule broken through the same entry point. That is all a tag says; a different bug behind the same entry point shows up at the fix (rounds).
A tag is a keyed hash of the rule and the entry point, under the program's key, which only the enclave holds.
At a proposed fix, the keeper re-runs every paid proof (the release gate holds only if all of them fail against it) and every duplicate and gap claim (the fix may promote them).
OutLayer's attestation names the pinned verifier wasm and the sha256 of the request near-antibody sent and of the output it received; NEAR's DCAP verifier accepted the run's Intel TDX quote over it. Instant payouts need it.
Holds if: the run is attested: its quote was verified on NEAR
Zero-knowledge proofconfirmed on NEAR
A zero-knowledge proof of the verdict
near-t2 verified an SP1 proof that the verifier, re-run on the enclave's exact input, gives the enclave's receipt, kind and rules, and compared every bound value with near-antibody's records. It claims no privacy (anyone can open the proof it re-runs: disclosed, or under a published or public content key), and no payout depends on it.
fa62fbea3e9e…, the proof's receipt sha256: fa62fbea3e9ea80fc38a8c5411b7041948e26add0c98efe501eafeec318d7c98
equals the enclave's
Kind
Violation
equals the enclave's
Slots (severity)
0 critical
equals the enclave's
A different request, run, blob, base, manifest, anchor, program, verifier or severity is rejected outright; a different receipt, kind or slots is recorded as a contradiction.