14 lines
1.7 KiB
JSON
14 lines
1.7 KiB
JSON
{
|
|
"actor": "final-controls-verifier",
|
|
"basis": "3c54cfd9a34913e0725e7eaa9f5ee04243a1a60d6edb10d0e2971e02dbf0a5d4",
|
|
"contract": "controls-bootstrap",
|
|
"evidence_digest": "7d2397a59ccba7abb61316ff8231eebe0547e563d4d9abb5d432ceb5fd828de2",
|
|
"explanation": "Scoped integrated pass for the sole control-tests criterion only. Independent reproduction from /tmp/opencode/sots-final-re ran exactly the declared 37 tests: 37/37 passed in 6.963s, with all expected lifecycle, identity, checkpoint, evidence/source-binding, lease/WIP, runner completion/config-event branches and the old-valid-recovery/resolution-only positive states named in the evidence. Held-out negative control used a nested duplicate JSON key (inputs) outside the declared evidence workload; read_json rejected it with ControlError. Actual source state was independently rehashed, not inferred from counters: paired engine digest ccd8e02083e8d2e2b3e97976ace2273c8f924dfc02a39e919004eaf3544c50fd and RE digest 6696fd5201e144843617cbf6d78b41b5287ad5dcc9fa1e8aaa861d52b64e72e8 match integrated evidence, with expected HEADs/common dirs. All declared binary/input/outcome SHA-256 identities match canonical evidence. No original executable, game state, RNG stream, original-assisted path, or full replay is in this tooling-only contract; those axes are unclaimed, not accepted. Residual: this attests tests/control instrumentation and exact package only, not security/authentication or universal engine correctness.",
|
|
"model": "openai/gpt-5.6-terra",
|
|
"role": "verifier",
|
|
"schema": "sots-verdict/1",
|
|
"session": "run-97066391080a3d06dce27a80",
|
|
"source_bindings_digest": "2b1bb449c29a7c7ad93a2ef9a5afaf06db54720abb58b5a249a20d9ab1317cb8",
|
|
"timestamp": "2026-09-09T22:35:28.183478+00:00",
|
|
"verdict": "pass"
|
|
}
|