docs(integration): retain finite same-goal worker evidence scope

This commit is contained in:
Agent Workbench Dev
2026-11-12 01:46:46 +08:00
parent 79e9f15f89
commit 211117e06d
7 changed files with 266 additions and 0 deletions
+2
View File
@@ -1,5 +1,7 @@
# Agent Workbench项目执行规范
2026-11-12 新协调来源只读登记在[同Goal工件连接记录](docs/integration/MNEMO-GOAL-WORKER-READONLY-2026-11-12.md):Mnemo指定旧AG42aacbd1/b330b966有限连接/显式complete已过,113=19+94、独立69及31隔离HTTP仅按来源回执范围,原9/build/57未重跑,本增量模型0;4321原观察404、未载入,生产者/完整指纹/派发接收/父交付/语义未验。AD与M4三工程分别登记,旧AG异常/Purge未验和Claude现有两403/unknown及541/41封存不改签。本消息不请求新执行或源线程回复;本次不读Key、不查正式服务、不重跑/重派或新增Claude/Purge,仅本地登记来源,后续开发阶段再按既有授权增量推送。
2026-11-12 本轮提交推送完成:修后Claude驱动97f132d3/验收5ca65a45已普通增量推送并与远端/fetch逐SHA一致,见[回执](docs/gitatlas-claude-proxy-correction-push.json);完成回执与接续最后另提交核最终HEAD。原M4收尾c8c36c27也已推送,两根541/41封存保持;当前Claude上游403客户端分组权限仍未通过,不追加本轮两请求。下一M8新原版Purge独立冻结,历史异常/unknown/真实预留/Mnemo语义/手机/长期边界保持。
2026-11-12 修后Claude独立两预检结束:目录完整14型号/选定在内;两新请求均上游403 permission_error“分组仅允许Claude Code官方客户端”,无响应型号/SSE/usage,用量/费用null、工具执行0。见[修后验收](docs/QA-CLAUDE-PROXY-CORRECTION-2026-11-12.md)。原58298行/137导出/418请求/5576工件及当前七连接/池/选择/配置/凭据行SHA一次exact,41成员36284086字节新封存SHA9b8a1cfc通过;原M4的40请求/541清单保持。不追加或重派两请求,不回退人类更正或冒充Claude Code,供应商须另开放标准Messages客户端。当前验收提交/普通推送收尾,产品未改不重发行/重启;M8原版Purge另冻,全部历史unknown保留。
@@ -121,6 +121,8 @@ Mnemo/其他产品从双方实际原件契约实现导出、入站委托、归
2026-11-11流式依赖登记:智枢NFD/ISS-041已部署明确failed分支的已有事件尾页排空,抛原错、不伪造成功、不重派;105回归/14部署等按来源结果,unknown/旧CT03及质量根因不改签。EQD已结束且不采用当前35B32K执行候选,严格7/8及原时延保留。见[只读登记](integration/MNEMO-NODE-FAILURE-TAIL-READONLY-2026-11-11.md);本轮无运行时复验或Workbench程序/模型变更,持续公网/手机/P8/真实收益边界与M4顺序保持。
2026-11-12 M7来源补充:Mnemo指定旧AG42aacbd1/b330b966同Goal原件连接和显式完成检查点有限范围完成,113相关/69独立/31隔离HTTP与原9/build/57复用单列,0模型。常驻4321原只读404/未载入,生产者认证/完整指纹/派发接收/父交付/语义仍未验;见[只读登记](integration/MNEMO-GOAL-WORKER-READONLY-2026-11-12.md)。原AD、M4三工程与该旧AG分开,不据此重派、重验Claude403或启动Purge,也不代签生产语义和收益。
### M8:真实收益单独证明
缓存指标取上游实际read/write/uncached;区分新增文件读取、内容变化后的必要回读与同SHA/同范围重复读取。提醒、经验、角色分别固定来源和开关,比较工程成功率/质量、tokens、时延及原失败,不把工具进展观察当业务交付。
@@ -0,0 +1,241 @@
{
"schemaVersion": 1,
"recordedAtUtc": "2026-11-11T17:44:27.815Z",
"recordedLocalDate": "2026-11-12",
"timezone": "Asia/Shanghai",
"scope": "Read-only public source metadata and named result receipts; current product phase and existing authorization remain independent.",
"sourceThreadId": "01a1ae3d-66a7-78b1-945e-b0c1d42c28ae",
"sources": [
{
"path": "D:/Ai-project/Mnemo/docs/integration/2026-11-12-goal-native-worker-checks-review.md",
"copyPath": ".qa/coordination-readonly-2026-11-12-goal-worker/review.md",
"bytes": 6227,
"sha256": "71f24c31ed4c11cd931d338fa1ccf3eaa1953fcf194b6006d0780a0dc1eb137d"
},
{
"path": "D:/Ai-project/Mnemo/docs/integration/goal-native-worker-checks-validation.json",
"copyPath": ".qa/coordination-readonly-2026-11-12-goal-worker/source-validation.json",
"bytes": 13905,
"sha256": "2c5a6a9058dd3140d609dc200f83506b5a7ab08ada864d7c52e54b33418b696d"
},
{
"path": "D:/Ai-project/Mnemo/docs/integration/coordination-2026-11-11T1729.json",
"copyPath": ".qa/coordination-readonly-2026-11-12-goal-worker/coordination.json",
"bytes": 32164,
"sha256": "6768561acf99ded187af307957f07d1c7e731d1e412cddb075745f70f0c1443a"
},
{
"path": "D:/Ai-project/Mnemo/output/goal-native-worker-checks/2026-11-11-user-continue/independent-http-FVYkLu/result.json",
"copyPath": ".qa/coordination-readonly-2026-11-12-goal-worker/actualEvidence.json",
"bytes": 39370,
"sha256": "41d994a81a2d0b8778053884c19d8d699d834ad41e16235e47630553dd5a3a8a"
},
{
"path": "D:/Ai-project/Mnemo/output/goal-native-worker-checks/2026-11-11-user-continue/tests-07-nAJKiX/result.json",
"copyPath": ".qa/coordination-readonly-2026-11-12-goal-worker/relatedNewTestRun.json",
"bytes": 2854,
"sha256": "34b3f7f981ecca566f38fb097ce18090aac47e18bebdeb04b52d983d1dbcdf7a"
},
{
"path": "D:/Ai-project/Mnemo/output/goal-native-worker-checks/2026-11-11-user-continue/goal-regression-01-lmA40f/result.json",
"copyPath": ".qa/coordination-readonly-2026-11-12-goal-worker/relatedRegressionRun.json",
"bytes": 2951,
"sha256": "f78243a86a12484996ad2df600b1ba0a43ce4a11a4f17d6cf42b0bdab4bbee3c"
},
{
"path": "D:/Ai-project/Mnemo/output/goal-native-worker-checks/2026-11-11-user-continue/types-07-iSdiNm/result.json",
"copyPath": ".qa/coordination-readonly-2026-11-12-goal-worker/typecheckRun.json",
"bytes": 2703,
"sha256": "7c366f9ee110cb2d5faa8f2116d0cb5a4aa62bd5c93f1418acb6e43a2ff7de48"
},
{
"path": "D:/Ai-project/Mnemo/output/goal-native-worker-checks/2026-11-11-user-continue/integrity-final.json",
"copyPath": ".qa/coordination-readonly-2026-11-12-goal-worker/finalIntegrity.json",
"bytes": 58872,
"sha256": "4ae23efff41de7a8030fcc747e8c0c43552bec0ed6b2a4ba7d2aafd991e5e244"
},
{
"path": "D:/Ai-project/Mnemo/output/goal-native-worker-checks/2026-11-11-user-continue/current-host-readonly-probe.json",
"copyPath": ".qa/coordination-readonly-2026-11-12-goal-worker/source-host-probe.json",
"bytes": 335,
"sha256": "2a213a12bc51a2fcc72a6b45fbdddeaf7b7e8af8218357d286f313c58fa21a17"
}
],
"claude": {
"connectionId": "9ae348a5-d0a6-42ed-8664-0da5f9effe34",
"connectionRevision": 9,
"credentialRevision": 2,
"catalogModels": 14,
"selectedListed": true,
"requests": 2,
"unknownRequests": 2,
"actualTotalTokens": null,
"actualTotalCost": null,
"upstreamHttpStatus": 403,
"upstreamType": "permission_error",
"message": "该分组仅允许使用 Claude Code 官方客户端。 This group is restricted to Claude Code clients (/v1/messages only)",
"textTaskId": "22f44c69-5b5e-4d23-9298-9207e7550e34",
"toolTaskId": "5decb3cb-ea08-4d14-aa28-2f697c37ebb0",
"completeHttpErrorAndUnknownUsageSeparate": true,
"oldMismatchResponses": 2,
"oldCatalogAbsenceRetained": true,
"modelCompatibilityQualified": false,
"repeatRequested": false,
"newScopeOnlyAfterActualClientPermissionChange": true
},
"mnemo": {
"status": "listed_finite_existing_worker_goal_scope_passed",
"nativeIdentity": {
"projectId": "4070c3e0-a292-45ba-bdbc-bcb4160f0fd1",
"sessionId": "91d910d3-7849-4f43-a0d5-1cfa019c34d7",
"taskId": "42aacbd1-f1fd-418c-80a9-9e1af9ff42c8",
"runId": "b330b966-848a-496b-a232-6b4438383ccc",
"workspaceId": "11771ddf-9dc1-4ba4-987f-34cc3d80c2ce",
"workerId": "2539daaf-ae13-4506-ac95-227032f606b0",
"baseCommit": "c213ad68236af508bf9b31e9739483caeda4ef0c"
},
"newChecks": 19,
"existingRelatedChecks": 94,
"relatedChecks": 113,
"independentChecks": 69,
"independentHttpRequests": 31,
"originalCodeReexecuted": false,
"originalTestsPreviouslyPassed": 9,
"originalIndependentChecksPreviouslyPassed": 57,
"originalModelRequests": 8,
"originalReportedTotalTokens": 109888,
"originalPriceVerified": false,
"modelCallsThisSourceIncrement": 0,
"readOnlyInspectCalls": 3,
"closedLoopbackHosts": 2,
"closedStoreCopies": 11,
"frozenInputsUnchanged": 147,
"initialFailedValidationAttemptsRetained": 11,
"sameGoalFiniteScopeCompleted": true,
"explicitCompletionCheckpointVerified": true,
"completeW2": false,
"producerAuthenticated": false,
"nativeFullSourceFingerprintVerified": false,
"parentDeliveryVerified": false,
"semanticQualityVerified": false,
"nativeDispatchReceiveQualified": false,
"normalHostRouteAvailableAtSourceObservation": false,
"normalHostReloaded": false,
"sourceHostProbe": {
"path": "D:/Ai-project/Mnemo/output/goal-native-worker-checks/2026-11-11-user-continue/current-host-readonly-probe.json",
"copyPath": ".qa/coordination-readonly-2026-11-12-goal-worker/source-host-probe.json",
"bytes": 335,
"sha256": "2a213a12bc51a2fcc72a6b45fbdddeaf7b7e8af8218357d286f313c58fa21a17"
},
"rawProbeFieldNames": [
"status",
"url",
"method",
"at",
"modelCalls",
"body",
"mutations",
"authenticated"
],
"historicalTaskRedispatched": false,
"sourceNext": {
"id": "U2-W2-next-dependency-supported-actual-task-review",
"status": "no_new_dependency_ready_task_selected_by_this_increment",
"implementationStarted": false,
"scope": "Only review newly available tasks/adapters/resources in the existing W2 plan; do not repeat this finite existing-worker task.",
"completionCondition": "核对新增、明确登记且未消费的实际任务,固定其事前判据与原生适配/manifest/资源/预算/原键;只实施已授权且依赖具备的部分。常驻服务载入需服务变更新授权,真实模型/embedding/付费/权重仍关闭。没有新具备依赖的待办就保持现状,不重复原九项/build、本地摘录或C5o~C5y。"
}
},
"counting": "113=19+94; independent69 includes selected-pin checks and is not added to31 HTTP or3 source inspections. Original9/build/57 and8 model requests are historical; this source increment made0 model calls.",
"taskSeparation": {
"adTaskId": "6be88bee-1995-424a-b15a-2e0e54207d1a",
"agTaskId": "42aacbd1-f1fd-418c-80a9-9e1af9ff42c8",
"m4TaskIds": [
{
"kind": "node",
"taskId": "8249fea9-3613-4830-b016-53446019c476"
},
{
"kind": "react",
"taskId": "dddcbdf2-588d-4fee-a26c-236e110a925d"
},
{
"kind": "python",
"taskId": "875afb7b-ea0b-41bf-8434-5897f5c800f0"
}
],
"allDistinct": true,
"oldAdInternalReportInProgressRetained": true,
"oldAgNetworkUnknownAndPurgeBenefitRetained": true
},
"sealedWorkbench": {
"originalM4Requests": 40,
"originalM4UnknownRequests": 0,
"originalM4Files": 541,
"originalM4ManifestSha256": "558c9b548559798279d7f8e61deb3b7344920753a53dbf4e84eec374c6f3a3df",
"correctionFiles": 41,
"correctionManifestSha256": "9b8a1cfcc4a8842e5c751990fdfc4a274831bf4dcafe73c557514c992490f2e5",
"rootsWritten": false,
"rawMembersReopened": false,
"verificationScope": "Existing public indices and manifest metadata only in this coordination; previous member verification remains separately recorded."
},
"ownReferences": [
{
"path": "docs/QA-CLAUDE-PROXY-CORRECTION-2026-11-12.md",
"bytes": 4881,
"sha256": "32087435370b79aa0939ffec16d7bd99edf41e3a9d55fc8b362c7645617eee8f"
},
{
"path": "docs/claude-proxy-correction-2026-11-12.json",
"bytes": 14308,
"sha256": "7c4cbc3f2897584f5601c846218e3713fb8f89a76b24f9a10412a9e192dfafab"
},
{
"path": "docs/claude-proxy-correction-seal-2026-11-12.json",
"bytes": 958,
"sha256": "9747526c8eab2c393069143656213a81b96091cdd80e7d8e69a90c99a0e298fc"
},
{
"path": "docs/QA-M4-LIVE-BASELINE.md",
"bytes": 6286,
"sha256": "ca2b27789a52060fa6bc2f273af48108f7eb610485e891bd727e48181bd4c75d"
},
{
"path": "docs/m4-final-2026-11-11.json",
"bytes": 167824,
"sha256": "e0ea050f24774d61c190d400a1e7bc6dcdb233e4c2d3076e10928dd4dd7046c3"
},
{
"path": "docs/m4-live-seal-2026-11-11.json",
"bytes": 622,
"sha256": "c5ac5623023f0a7f89a2c8f95dbb1e19b76103e39b8f734f21c990f878e45cea"
},
{
"path": "docs/integration/WORKBENCH-AD-NODE-READONLY-SOURCE-2026-11-11.json",
"bytes": 25974,
"sha256": "a0a1fdb246a07b3a64c160c307545bd1e00c2f182adcebaa50ab9f6f0f30c9f6"
}
],
"verification": {
"ninePublicCopiesExact": true,
"fivePinnedSourceReceiptHashesMatched": true,
"related19And94ReceiptCountersMatched": true,
"independent69ReceiptPassedMatched": true,
"typecheckExitCodeMatched": true,
"selectedAGIdentityMatchesPriorSource": true,
"distinctFromADAndM4": true
},
"boundaries": {
"modelRequests": 0,
"engineeringExecutions": 0,
"externalHttpRequests": 0,
"credentialsReadOrResaved": false,
"otherProjectSourceWritten": false,
"formalServiceReadOrRestarted": false,
"oldSamplesReplayed": false,
"messagesSent": 0,
"subagentsOrAutomationsCreated": false,
"authorizationExpanded": false,
"sourceAuthenticationDatabasesPermissionsAndModelPreferencesInherited": false
}
}
@@ -0,0 +1,15 @@
# Mnemo 同 Goal 旧 Worker 工件连接:新增只读结论
2026-11-12(上海时间)接收三项目协调来源,按当前记录范围只读核对;机器索引:[MNEMO-GOAL-WORKER-READONLY-2026-11-12.json](MNEMO-GOAL-WORKER-READONLY-2026-11-12.json)。九份公开来源/结果回执原字节副本在 `.qa/coordination-readonly-2026-11-12-goal-worker/`,五份发布的回执bytes/SHA、19+94检查计数、69独立结果及类型退出码已核对。未执行来源检查、HTTP、模型请求或工程,也未查询/重启正式服务、读Key或回复来源线程。
Claude修后现有两用例已经结束:revision9/credentialRevision2,完整14型号目录包含claude-opus-5-5;文本Task22f44c69、工具Task5decb3cb各一请求,均上游403 permission_error,仅允许Claude Code官方客户端。完整188字节HTTP错误与usage未知分别成立,实际tokens/费用null;旧两次自报GPT、旧目录缺项和原M4四十请求/541成员封存均保持。当前只核已有公开索引及清单元数据,没有重开原封存成员、再次验收或重POST。本协调不请求重试/Purge;实际标准Messages客户端权限变化后也需按当前用户授权另冻新用例。
Mnemo已完成人类指定旧AG ba-off的有限同Goal工件连接/显式complete检查点:Task `42aacbd1-f1fd-418c-80a9-9e1af9ff42c8`、Run `b330b966-848a-496b-a232-6b4438383ccc`、子工作区及HEAD c213ad68与原登记逐字段一致。相关113=新19+原94;独立69包含选定pin核对,31次隔离HTTP、3次只读inspect和两个关闭宿主单列,不能累加成更多独立测试。11次原失败保留。原九项/build/57结果未重跑,旧8请求/109888报告tokens及未核实价格保留,本增量模型调用0。
这证明既有原件连接和隔离本地服务守卫的有限范围。常驻4321的新入口只有来源当时未认证GET的404观察,未载入/重启本包;本轮没有重新探测。生产者认证、完整原生来源指纹、原生派发接收、父交付/完整交付报告、语义质量和完整W2仍未验证,文件复核也不提供跨进程原子快照。Core SQLite权威、147冻结与原质量失败保持。
原AD的Task6be88bee、这个旧AG与M4三个新工程均是不同对象,不能把一方结果补签到另一方。原AD内置报告in_progress、原AG三次网络失败/unknown及唯一ba-off工程成功、Purge收益未验证等原结论不改签;当前M4三工程成功也不替代Mnemo生产或记忆语义验收。
后续M7原生集成引用这份限定证据;M8新原版Purge仍按独立冻结计划。三产品认证、数据库、权限、模型偏好及质量门槛不互相继承。本消息没有新增执行授权,本次仅登记来源并沿用已有本地小提交纪律,待后续开发阶段统一增量推送。
来源:[Mnemo回顾](D:/Ai-project/Mnemo/docs/integration/2026-11-12-goal-native-worker-checks-review.md)、[机器索引](D:/Ai-project/Mnemo/docs/integration/goal-native-worker-checks-validation.json)、[本轮协调记录](D:/Ai-project/Mnemo/docs/integration/coordination-2026-11-11T1729.json)。此前[NRO只读登记](MNEMO-NRO-READONLY-2026-11-11.md)与Workbench原验收不修改。
+2
View File
@@ -1,5 +1,7 @@
# 汇总发现
2026-11-12 Mnemo来源指定旧AG的同Goal工件连接已有限完成;113=19+94、独立69/31隔离HTTP分别核回执,原9/build/57及8请求109888tokens不重复计。本轮只读九副本/五发布回执SHA,4321的404是来源旧观察,不称当前生产接通;生产者/完整指纹/派发接收/父交付/语义未验。Claude403两请求/usage未知和原541/41清单保持,见docs/integration/MNEMO-GOAL-WORKER-READONLY-2026-11-12.md;不重派或补签AD/M4。
2026-11-12 修后Claude验收5ca65a45和驱动97f132d3已普通推送、远端/fetch逐SHA一致,见docs/gitatlas-claude-proxy-correction-push.json;原M4收尾c8c36c27也已完成。541旧封存/41新封存保持,Git不含原件/Key/程序包。目录成功与403客户端权限失败并列,真实usage/费用null和全部历史收益unknown保持。
2026-11-12 修后Claude的目录已完整14型号且含选定,但两模型请求均403 permission_error,原文明确分组仅允许Claude Code客户端。实际/v1/messages、x-api-key占位、stream及max_tokens128000字段完整,响应188字节JSON完整,不能因.sse后缀当流式通过;无型号/usage,实际tokens/费用null。规划requestId/物理wireAttemptId从原账本登记,可选requestKey空字段保持,不重派。原状态与当前七连接一审通过,41新封存及旧541保持;见docs/QA-CLAUDE-PROXY-CORRECTION-2026-11-12.md。需要供应商普通API客户端权限,不以目录成功认证权重或收益。
+2
View File
@@ -1,5 +1,7 @@
# 统一执行进度
2026-11-12 三项目新增只读来源核对并登记:九公开副本、五已发布回执bytes/SHA、19/94/69计数/类型退出码和旧AG原身份通过;docs/integration/MNEMO-GOAL-WORKER-READONLY-2026-11-12.md保留有限同Goal/显式complete与尚未载入/认证/父交付/语义边界。0模型、0工程、0HTTP或正式服务操作,不回复来源、不改原封存。按既有本地小提交纪律收尾,待后续开发阶段增量推送;不因本协调新增Claude/Purge执行。
2026-11-12 本轮实际执行/验收/封存/推送结束:原M4 c8c36c27、修后Claude97f132d3/5ca65a45已与远端/fetch一致;docs/gitatlas-claude-proxy-correction-push.json与完成接续最后限定提交后核最终HEAD。541/41封存保持,Claude403未通过且两请求不追加;下一M8另冻,不重跑已过工程或改变旧异常。
2026-11-12 Claude独立修后范围结束:97f132d3驱动离线10项通过后一次冻结,目录14完整/选定在内,两新例各一请求均上游403“仅Claude Code客户端”,2unknown、费用null、0工具执行。audit-44ea125a一次保持原58298行/137导出/418请求/5576工件及当前七连接/池/选择/配置/凭据SHA;41成员36284086字节SHA9b8a1cfc封存复核,原M4 541清单保持。docs/QA-CLAUDE-PROXY-CORRECTION-2026-11-12.md记录原错误和完整身份,当前限定验收提交/普通GitAtlas推送收尾,不追加或重跑工程/Purge/重启。
+2
View File
@@ -1,5 +1,7 @@
# Agent Workbench:核心代理能力研究与产品计划
2026-11-12 接收只读协调,九公开副本/五发布回执SHA及指定旧AG身份核对;docs/integration/MNEMO-GOAL-WORKER-READONLY-2026-11-12.md登记Mnemo同Goal有限连接/显式complete、113相关/69独立/31隔离HTTP,生产/父交付/语义仍未验。AD、M4三工程分别保留;Claude两403/unknown和541/41原封存不变。本次不新增执行/查正式服务/源线程回信,只本地记录,下一开发阶段沿既有授权推送。
2026-11-12 本轮收尾完成:原M4 c8c36c27和修后Claude驱动97f132d3/验收5ca65a45普通推送后远端/fetch一致,docs/gitatlas-claude-proxy-correction-push.json与完成接续最后另提交核最终HEAD。541/41两封存保持,三工程成功及Claude新403分别成立;两请求不追加,下一M8新原版Purge另冻,历史unknown不改签。
2026-11-12 修后Claude执行和证据完成:目录完整14/选定在内,两新请求HTTP403 permission_error“仅Claude Code客户端”,0执行工具、2unknown,实际tokens/费用null。docs/QA-CLAUDE-PROXY-CORRECTION-2026-11-12.md与机器记录/41成员封存SHA9b8a1cfc通过;一次核原58298行/137导出/418请求/5576工件与当前七连接/池/选择/配置。当前限定验收提交及普通推送;原40请求/541清单保持,供应商权限须另解决,不追加请求或重跑工程/Purge,下一M8另冻。