{
  "schema": "agent-completion-repair-manifest.v1",
  "sourceSha": "75fb1c7a5473549983c16143e3cb15482c940d97",
  "generatedAt": "2026-10-16T13:22:22.547358+00:00",
  "rowCount": 63,
  "protocol": [
    "Unchanged frozen parser prompt, starter and hidden oracle; both actual CLIs. The user replaced the obsolete roster with five current inexpensive models, selected from live ladders and catalog added dates; historical runs remain in audit history. Every historical campaign and every new attempt remains available.",
    "Both agents request temperature zero and retain fixed 240-second, per-request and cumulative token limits. GPT-6 Luna uses Responses with the same provider-default reasoning in both arms; its unsupported custom sampling is omitted upstream, as required by the vendor. No silent model substitution or hidden-check feedback.",
    "Generic product changes remove redundant read and completion turns, bound repeated self-repair and provide an explicit text tool protocol for models without native calls. Protocol differences are disclosed in receipts.",
    "The independent hidden scorer runs only after candidate execution. It supplies no examples, diagnostics or expected outputs to candidates or repair decisions.",
    "Latest scheduled attempt is displayed even if worse or incomplete; no best-of selection. Causal apparatus retries remain in history.",
    "Other coding and Blueprint families retain prior measured receipts; they were not rerun in this parser campaign.",
    "Synthetic single-task fixtures, shared hardware and one trial per arm. No statistical or general superiority claim; canonical benchmark-store admission remains unverified."
  ],
  "originalManifest": "/benchmarks/agent-comparison-20261015/manifest.json",
  "originalFixtures": "/benchmarks/agent-comparison-20261015/fixtures.json",
  "receipts": [
    {
      "file": "20261016T122718Z_aider_byok_openai_gpt-4.1-mini_c41098.json",
      "sha256": "91aa6f55e2ffebb3ba5a3cd33039ecedfe5dc7894b2b915627d17f20ec26b8f0"
    },
    {
      "file": "20261016T122849Z_aider_byok_deepseek_deepseek-flash_0893d1.json",
      "sha256": "ffa441f980ca69b0aa557a57cc0e2ddb3a3cf78c62a05bd022ab804aac9f9e10"
    },
    {
      "file": "20261016T122849Z_aider_byok_qwen_qwen3-coder-flash_0cda6e.json",
      "sha256": "7f7f6433614075bd2619f10abd5560fb5eee9ea058b737a63a6815d4aad9de7c"
    },
    {
      "file": "20261016T122849Z_aider_byok_zai_glm-5.3-flash_417eeb.json",
      "sha256": "9dbeebc6746a9802cbf6369b2973acaa3a8dedb034494061d4e7fc5ef424e7f1"
    },
    {
      "file": "20261016T122858Z_aider_byok_google_models_gemini-2.5-flash_5a87e5.json",
      "sha256": "edda7d805e309dd75aab1a6c0f036c6881ba64afdb1e8ffb8df121b23da7a4e5"
    },
    {
      "file": "20261016T122901Z_aider_byok_qwen_qwen3.8-flash_c9d466.json",
      "sha256": "d9ecb605f8529d904cfd8a06d86773b8d3e771027603b745c8215460e1107e1b"
    },
    {
      "file": "20261016T122917Z_aider_byok_google_models_gemini-3.8-flash_43aead.json",
      "sha256": "2c6958a63c5416cf2eb426070ea3f209372c1c2e7d1e6995047496c0975079e4"
    },
    {
      "file": "20261016T123139Z_verdict-code_byok_openai_gpt-4.1-mini_154f5e.json",
      "sha256": "9bc469efec42f173eab094f18cb480c0efd176c9b66f1a7dd1ebe178dab7f927"
    },
    {
      "file": "20261016T123350Z_verdict-code_byok_openai_gpt-4.1-mini_9b3037.json",
      "sha256": "06f799e4cbe07f1ddacc3e2f7de8a3f70c57996e136116d92ee7a2a1ea001e80"
    },
    {
      "file": "20261016T123447Z_aider_local_lmstudio_verdict_verdict-4-35b-toolterse-v5-mlx_248d97.json",
      "sha256": "0efd6caef2036da20fe35360034b0a8d45ae6465b9b81744b925283d43aab9b1"
    },
    {
      "file": "20261016T123456Z_aider_local_lmstudio_qwen_qwen3-coder-30b_95980d.json",
      "sha256": "661a45d83ce13c6d956d91de68bfe5d089037145685e3648c7525934e3d1c4f0"
    },
    {
      "file": "20261016T123506Z_aider_local_lmstudio_openai_gpt-oss-20b_55fecb.json",
      "sha256": "a2be895e209efa46ff66da00b260104657dcdd25f0515423687878ca8a9009ff"
    },
    {
      "file": "20261016T123533Z_aider_local_lmstudio_google_gemma-4-26b-a4b_9410a4.json",
      "sha256": "dcdfa0157b79b2b1d8bbb67bd7d1be967ccd05d8644e0172bb93a9f5628c8af7"
    },
    {
      "file": "20261016T123611Z_aider_byok_qwen_qwen3.8-flash_5dd683.json",
      "sha256": "65639932fc82df03e10a9a638d91a2e2ea51c36ae536ebf7e5f4594cd0af7229"
    },
    {
      "file": "20261016T123719Z_verdict-code_byok_deepseek_deepseek-flash_0e44de.json",
      "sha256": "c2affa6711469d8ec8af5409fcba1691536ce87c01129f0d56e3ad289f9ff7d3"
    },
    {
      "file": "20261016T123719Z_verdict-code_byok_openai_gpt-4.1-mini_123c40.json",
      "sha256": "becb6dd70e9749cc4086c0c31abb82509a9200ccda18e85f8c259ed1323a98a7"
    },
    {
      "file": "20261016T123719Z_verdict-code_byok_zai_glm-5.3-flash_0f8542.json",
      "sha256": "333dc60f384b4d477ef4858b444e99b43d271d4162261fce6b9ef88ebdf48e16"
    },
    {
      "file": "20261016T123732Z_verdict-code_byok_qwen_qwen3-coder-flash_62aae4.json",
      "sha256": "8e34e4cccf3507fe789277a03a5f0e394b53050a561678123ec385d9f77ad7e1"
    },
    {
      "file": "20261016T123739Z_verdict-code_byok_google_models_gemini-2.5-flash_049660.json",
      "sha256": "f4ce9537ce987656462271a8df213d4f4e39be21a636492e67d1c500c8ff820d"
    },
    {
      "file": "20261016T123745Z_verdict-code_byok_qwen_qwen3.8-flash_b014b3.json",
      "sha256": "96e889cf05856e646837883bad53dee18c5f805d5ca5f6e6ce9bedc6e58cb2fc"
    },
    {
      "file": "20261016T123753Z_verdict-code_byok_google_models_gemini-3.8-flash_30a6fd.json",
      "sha256": "0ea2fff765916333985aa855a4d0e70b971530dcc5cee3188bd779adfce6f980"
    },
    {
      "file": "20261016T124009Z_aider_local_lmstudio_qwen_qwen2.5-coder-1.5b-instruct_0f192d.json",
      "sha256": "87243acfc24f108c6add136b8bae49fa7d1399e38da246b951e3af7bab260c7b"
    },
    {
      "file": "20261016T124645Z_verdict-code_byok_anthropic_claude-haiku-5-5_677cef.json",
      "sha256": "25fb6a7732e3d84fc2e8626ced9129690f8ca613dd54ea4b32931f59afbe24a1"
    },
    {
      "file": "20261016T124926Z_aider_byok_zai_glm-5.3-flash_170c14.json",
      "sha256": "47ea66291c6d57323263cabaef579b8236c58005a51726ff362c1561d521a734"
    },
    {
      "file": "20261016T124951Z_aider_byok_anthropic_claude-haiku-5-5_7731ed.json",
      "sha256": "9bf34772cc284e0589fce73d48f9982f684f9eb7c7c0a49b3a9dfa6cfe3abe31"
    },
    {
      "file": "20261016T124951Z_aider_byok_openai_gpt-6-luna_fd7473.json",
      "sha256": "6d7cb42c990c7fc56f7afd2a28d46b7dee6fa4afb3da4b1ad5ce55e821696e34"
    },
    {
      "file": "20261016T124951Z_aider_byok_qwen_qwen3.8-flash_d64870.json",
      "sha256": "b8df829550cd71246a37ef333430a19cd313f107dd3c287c0858c1a1026d4d2e"
    },
    {
      "file": "20261016T124951Z_aider_byok_zai_glm-5.3-flashx_ffee9b.json",
      "sha256": "30e692009c7ec7cce6cf433ae1ec6936473b226389f16de58efe4be7677b06dc"
    },
    {
      "file": "20261016T124951Z_verdict-code_byok_openai_gpt-6-luna_ebbc95.json",
      "sha256": "47aed478ad9d2a44c3647469147107ce5d987b5b9a368af8dab7e154a14d69ee"
    },
    {
      "file": "20261016T124951Z_verdict-code_byok_qwen_qwen3.8-flash_029acf.json",
      "sha256": "7793f8b2a547a26f137821a0282cefbdf7725300383856ceba0b500845ad5f24"
    },
    {
      "file": "20261016T124951Z_verdict-code_byok_zai_glm-5.3-flash_58dcd7.json",
      "sha256": "9c387f1f12135ee07e593a7e9b806aefdc7b839b663a124410b8229d1209d37b"
    },
    {
      "file": "20261016T124951Z_verdict-code_byok_zai_glm-5.3-flashx_5f0d6b.json",
      "sha256": "a64b04b5dc116d360b2e967fde3e71848559c0aa9bf693f3fbe1abc13e6291fc"
    },
    {
      "file": "20261016T125355Z_verdict-code_byok_anthropic_claude-haiku-5-5_62d411.json",
      "sha256": "8d2d44962b181073ca850802bbc0cb28b4f872253256b0ce03b2c7f2b10e954e"
    },
    {
      "file": "20261016T130549Z_aider_byok_openai_gpt-6-luna_563ba0.json",
      "sha256": "be2a0ba5c50389ff2fbe580768a591c7e75a991682156f6cd39f425d7baa8f15"
    },
    {
      "file": "20261016T130549Z_verdict-code_byok_openai_gpt-6-luna_4f07c2.json",
      "sha256": "f93c191912ad9014acc71c6c9e724681bfeaf42d6c7a4bd9258f1b31d2501d3d"
    },
    {
      "file": "20261016T130549Z_verdict-code_byok_qwen_qwen3.8-flash_fee25f.json",
      "sha256": "cd1f56a19cb13e485cd332a687a1fc5915a467e4405531a054a8dda64204cc9a"
    },
    {
      "file": "20261016T130549Z_verdict-code_byok_zai_glm-5.3-flash_9801e4.json",
      "sha256": "a81339c47cbb5efd79c31493e8e9a1a4a1d6a0b3985811c66d2e039752faffc7"
    },
    {
      "file": "20261016T131454Z_verdict-code_byok_qwen_qwen3.8-flash_7fd4b1.json",
      "sha256": "a1280b1dc132e169ae189be033d9c45798b8f88ceb8154b223c7997ac0efae9a"
    },
    {
      "file": "native-devin-extra-regression-repair.json",
      "sha256": "9eaf99ab4c55338f8c362ad868389487e09b0db2560a03b80a5cb0125ddb1ca0"
    },
    {
      "file": "native-claude-extra-regression-repair.json",
      "sha256": "d18a40c59a1d0938abf741b25eb8600f427f6850c7e6be9120021151de20af4c"
    },
    {
      "file": "native-codex-extra-regression-repair.json",
      "sha256": "19336061a466b63fb85f69f204370237156bba0efad706dcfd1250dde8a7e623"
    },
    {
      "file": "native-devin-blueprint-explicit-dependencies.json",
      "sha256": "dabb9d6c727b4e4b720c7e3709779f9258e11c195008974545bd6f0a760d4499"
    },
    {
      "file": "native-devin-extra-dependency-scope.json",
      "sha256": "8b2d9729d5b7096229f46211e20c2f4f3cff43f6b93a1f2ead9293dab1b4e90b"
    },
    {
      "file": "native-devin-blueprint-prose-scope.json",
      "sha256": "cf97b4a179a72b6abf3cbf642cc0df438622ca03ffb725f3182043223b5a65fd"
    },
    {
      "file": "native-codex-extra-dependency-scope.json",
      "sha256": "c1a8003e6d54e69e40beb681919c67e337d3c48a35cd58b5e1daff875db6b09f"
    },
    {
      "file": "native-codex-blueprint-explicit-dependencies.json",
      "sha256": "17e0ddb3393f3ecddee1f11d000b7328696ac41cd3880b16cafc434e82e3f6e2"
    },
    {
      "file": "native-claude-extra-dependency-scope.json",
      "sha256": "d3381d6271a9e6955417686bb8f70b19ef1eb919706aad5d75563f00ae7c40c6"
    },
    {
      "file": "native-codex-blueprint-prose-scope.json",
      "sha256": "b87343ba18559aba178875fccea778f9c6dd6e8cacad5311f2abed86f4dba044"
    },
    {
      "file": "native-claude-blueprint-explicit-dependencies.json",
      "sha256": "b4cc8521f0e50d675b133abbf994b285f8f84da9c0e234f947a3514d92ad056c"
    },
    {
      "file": "native-claude-blueprint-prose-scope.json",
      "sha256": "a7174dc00bcf24d00a697b10c6931f7d62c6476fbcdcfbf15473a335b2b0f927"
    },
    {
      "file": "native-antigravity-preflight.json",
      "sha256": "73b945b63ede07f6dd794cb22f92f1fbbf368688a55bdfb134495dd34c9a3c02"
    },
    {
      "file": "TEST-blueprint-planning-20261016T035726Z-blueprint-explicit-dependencies-blueprint.json",
      "sha256": "80e24add33bc5556f3c50adc64dd9fb35cb8ffc086ede02ecee49074ed178902"
    },
    {
      "file": "TEST-blueprint-planning-20261016T035733Z-blueprint-prose-scope-blueprint.json",
      "sha256": "940703a3e0cf8295c92e782a9d467f12ff693253dbe2cc8001cc6fb609f105f4"
    },
    {
      "file": "TEST-blueprint-planning-20261016T035735Z-blueprint-explicit-dependencies-raw-glm.json",
      "sha256": "bee96bc834af281e9e49631c500b3aeae4439fdeb183dc12559995f2f2debefb"
    },
    {
      "file": "TEST-blueprint-planning-20261016T035735Z-blueprint-prose-scope-raw-glm.json",
      "sha256": "dc7dd6c2da98310204b6c942972e1b7ed7a77b79f9a29c8d2f9e0a661f4d6e59"
    },
    {
      "file": "TEST-extra-dependency-scope-verdict-code-20261016T040002Z.json",
      "sha256": "957831f652e46a84b1401fdba0bd9f938b0b93a27ff9ffc22eccbcda658982de"
    },
    {
      "file": "TEST-extra-dependency-scope-verdict-code-20261016T040152Z.json",
      "sha256": "65db83ab8bbf028f46874db789e8ad322d932f015c8e87af65fb8461fce93eb5"
    },
    {
      "file": "TEST-extra-dependency-scope-verdict-code-20261016T040534Z.json",
      "sha256": "1297fed532cebdf61cd309e1764c00df8e454bd288725734fd713c9531b1a21b"
    },
    {
      "file": "TEST-extra-dependency-scope-verdict-code-20261016T042131Z.json",
      "sha256": "580bc969c0c7d5a61924e43a7e8ba1f2f28af2da6e66e54c93a6ca446029986d"
    },
    {
      "file": "TEST-extra-regression-repair-verdict-code-20261016T035725Z.json",
      "sha256": "d944229f81c3554fe07c7a1e6db793581252adbd27b90d8b763a82ec8fa80351"
    },
    {
      "file": "TEST-extra-regression-repair-verdict-code-20261016T040029Z.json",
      "sha256": "0f7f8b22fc4cb62037ea4fa2de701d0eb84beefe97de383bd58c881f20c30222"
    },
    {
      "file": "TEST-extra-regression-repair-verdict-code-20261016T040146Z.json",
      "sha256": "73129ebb06bb3e64fe59351336ca4aea1289e359ce5fea338aad62d79e81890c"
    },
    {
      "file": "TEST-extra-regression-repair-verdict-code-20261016T040304Z.json",
      "sha256": "fd9b130183d35f44179a4b55ccb54ad8c38af87e8de704e7733f7dc1577fc4a5"
    }
  ],
  "supportingFiles": [
    {
      "file": "captured-cost-accounting.json",
      "sha256": "317ef0507301acdf8e5a520c01cc0636302c8e09d479de6da228f17f5a3aa3be"
    },
    {
      "file": "final-qwen-causal-diagnostic.json",
      "sha256": "f5c4fd2de716ae4e554e1819a16e4cb888ba09b8fde710748a94c63ef9a7b83d"
    },
    {
      "file": "gateway-lifecycle-correlation.json",
      "sha256": "a78c6586db1f96a3ef606271607fc6b2307c47c209a003638c731cfce1412b04"
    },
    {
      "file": "gateway-runtime-proof.json",
      "sha256": "215cc89f7f171bdf87e312ce6dac910911c9d592b6852c426b51299d1b6abbe5"
    },
    {
      "file": "latency-audit.json",
      "sha256": "b56c5ed543bb2b74cb5e22dfdad40674a9a69c3e355e0540f6935be0ad479a46"
    },
    {
      "file": "latency-diagnostics.json",
      "sha256": "6577e5521e031f9ab72bf8bd45d97cd06881ad5a674420f5a556e4642bf3558e"
    },
    {
      "file": "model-selection.json",
      "sha256": "e6134279872dba35d64f7388a70a1da515817457f091a54f5a929df79adbf324"
    },
    {
      "file": "planning-rubric-v2.json",
      "sha256": "efad1e30bfa77262f70e44445ec3640106deb908ee2172c8d9edf073931dc6c8"
    },
    {
      "file": "preflight-diagnostics.json",
      "sha256": "fe541557d4a1e144e3d74ed453639aa4ee84d3e1ddb527774d77101935438141"
    },
    {
      "file": "preservation-audit.json",
      "sha256": "7024787ac7d9d63fe408d7eff890add08c3e6734b10622947503bfeebce70972"
    },
    {
      "file": "responses-runtime-proof.json",
      "sha256": "8edc3702aa5821b130a08c94126d8138030432f5c05749d83147368bac3680f5"
    },
    {
      "file": "root-causes.json",
      "sha256": "991039a559ea2bde78e851f28afafd54f94a2ed98ed6d99943621bc1ede17266"
    },
    {
      "file": "summary.json",
      "sha256": "dba0238d746182c8fce71f9d57ca6721ed43dca99c07dcc60f5aa14f54040bf0"
    }
  ],
  "originalSnapshotSha256": "379302059786c3766d7b3c91b8f63cf8e925109f14f6c77d293d1bb6a80f04af",
  "priorManifest": "/benchmarks/agent-comparison-20261016-harness-repair/manifest.json",
  "priorSnapshot": "/benchmarks/agent-comparison-20261016-harness-repair/summary.json",
  "priorSnapshotSha256": "e5d2cc49a571e32b674f3205465572157f7e42689f1b7e5f24b6dcdba38bf5fd"
}