docs: audit milestones, record confirmed fixes and gate the next development plan

This commit is contained in:
Agent Workbench Dev
2026-10-30 23:38:53 +08:00
parent 7b1f529ed6
commit 057d91f14a
48 changed files with 7181 additions and 28 deletions
+8
View File
@@ -1,5 +1,13 @@
# Agent Workbench 更新记录
## 1.12.1 · 2026-10-30
- 项目文件保护同时核对实际链接目标及父目录,拒绝指向Git/凭据的内部别名和Windows尾点/空格路径;多硬链接文件在读取、提案与自动规则前拒绝,普通项目文档链接与独立副本仍可使用。
- 模型选择与配置同事务保存;磁盘失败恢复已提交状态,整批导入不留下部分连接,失败预览可继续核对。
- 热升级保留原模型中心/导入器/凭据对象并更新方法;在途模型探针、排队与租约阻止冲突的配置变更或升级。
- 本地页面禁止嵌入,脚本和API限定同源;终端及Node/Python检查只接收明确的系统/运行环境,避免继承宿主未知凭据。
- 完成1.0–1.12里程碑与24项能力核对,记录[后续计划](docs/NEXT-STEPS.md);补丁独立验收、原失败与部署状态见[审查报告](docs/MILESTONE-AUDIT.md)及[1.12.1验收](docs/QA-1.12.1.md)。
## 1.12.0 · 2026-10-30
- 新增专业模型中心:多连接增改/启停、轮询/权重、Run内粘连、并发租约、限流/故障冷却和逐次请求账本;临时凭据重启后如实显示待配置。
+25 -25
View File
@@ -107,32 +107,32 @@
- C12/C13:成功准备Git预览后保存审查边界并停止模型;后续同消息工具不执行且协议完整。确认后内核完成原Task与报告,不继续请求模型,预览可修改/撤回。
- 192项回归和72组界面通过;合成模型子工作区真实3项失败→3项通过/build/Git报告通过。不可变包、原进程更新与真实配置模型结果见[QA-1.6](QA-1.6.md),完整三类工程及M2/M6能力继续验收。
| ID / 能力 | 1.0.0基础与缺口 | 后续产品目标 | 参考问题 / 模块 / 阶段 | 必须看到的验收证据 |
| ID / 能力 | 当前已交付能力与缺口(至1.12.0) | 后续产品目标 | 参考问题 / 模块 / 阶段 | 必须看到的验收证据 |
| --- | --- | --- | --- | --- |
| C01 需求与任务规划 | 有对话和有限步数工具循环,没有任务依赖图 | 固定目标、约束和完成标准;拆任务、判断依赖、接收追加要求与重规划 | CX-1、WB-1 / tasks / M1–M2 | 追加要求后仍保留原目标;任务完成由验收条件决定 |
| C02 执行状态与恢复 | 会话JSON保存,active/waiters在内存;重启不重放命令 | 独立Task/Run/Step身份、持久事件、检查点与未知结果核对 | CC-2、CU-1、DS-1 / storage、tasks / M1 | 在审批、调用、结果与落盘等切点中断;恢复不重复已完成副作用 |
| C03 工程识别与索引 | 提供有界文件清单/文本搜索,没有工程索引 | 识别语言、入口、规则、脚本、依赖和Git;增量索引;符号/引用与必要的语义检索 | CU-2、DS-3 / context / M2 | 对Node/React/Python样本定位实际修改入口;代码变动后索引更新 |
| C04 上下文与项目归属 | 有提取式消息投影、运行诊断和当前项目资料 | 为Run组合需求、规则、代码、错误与验证结果;引用来源;严格绑定项目 | CC-2、CU-1 / context、contracts / M1–M2 | 切换两个项目/任务后,错误与工具结果仍归属正确;大输出用工件引用 |
| C05 长期项目记忆 | 项目所属持久条目、来源/工程指纹、到期/归档、不完整扫描降级、有界且脱敏的请求接入与原快照已实现 | 自动提取/归并、长任务恢复包及真实模型收益继续验收 | CC-2、WB-1 / context、storage / M4 | 长任务压缩/重启后完成标准不丢;过期代码事实被更新或标记 |
| C06 工具、Skills与MCP | 自有工程工具/固定REA MCP;1.12源码支持项目Skills登记、启停/修订、按需读取与请求证据,见SKILLS.md | 动态MCP、schema与能力声明;插件安装/调用有来源和范围 | CC-1、CU-1、WB-2 / tools / M3 | 未知工具、错误参数、工具失联有原调用身份;新工具可按契约接入 |
| C07 执行控制 | 文件范围/基线检查、逐次命令确认与进程停止 | 操作范围、授权记录、预算、进程所有权、并发准入和取消传播;沿用用户已授权的范围 | CC-1、CU-1、DS-1 / tools、tasks / M1–M4 | 过期授权不能用于新Run;停止只处理所属任务;修改后的输入重新校验 |
| C08 精确编辑与多文件变更 | 完整替换单文件提案,有备份与冲突拒绝 | patch、创建/改名/删除、多文件ChangeSet、基线校验、隔离应用与可恢复合并 | CU-2 / workspaces / M2 | 跨4文件修改保留无关内容;外部改动引起冲突;部分失败有明确状态 |
| C09 用户工程Git | 本产品源码已建Git;界面内用户工程Git未实现 | 仓库/dirty状态、基线、任务分支/worktree、diff、审查后提交与合并 | CX-1、CU-3、DS-3 / workspaces / M2 | 保留用户已有修改;每个交付可定位commit;冲突合并可审查 |
| C10 长命令与开发服务 | 命令最长60秒,stdout/stderr分别截取32k字符 | 持久job、完整日志工件、心跳/退出码/超时、后台开发服务及所属进程树 | DS-3、CU-3 / tools、artifacts / M2 | 超过60秒的检查可观察和停止;失联结果标未知;重启不擅自再运行 |
| C11 真实检查与TestRunner | 可执行命令,没有结构化检查计划与报告 | 识别真实项目检查;计划绑定planId,执行返回jobId;保存退出码、测试数和报告 | DS-3 / tools / M2 | 构建失败、零测试、拒绝/取消均不算通过;结果绑定相同工作区与ChangeSet |
| C12 修复与代码审查 | 模型可根据普通工具输出继续,没有独立修复/审查闭环 | 从真实失败定位并有限轮修复;核对需求、diff、回归与风险;保留失败历史 | CX-1、DS-3 / tasks、workspaces / M2–M4 | 能复现的失败先出现再修复;实际重跑通过;审查不能用“模型说成功”代替 |
| C13 可交付成果 | 有提案/会话导出,没有工程交付清单 | 需求与结果对应、diff/commit、测试报告、运行方法、限制与工件引用 | WB-1 / artifacts / M2–M6 | 用户按交付说明能启动并验证;报告与实际文件、检查及commit一致 |
| C14 模型协议与路由 | 一个全局兼容连接;stream:false;基本usage记录 | ModelProvider能力探测、流式文本/工具参数、模型配置快照、按任务角色路由与可解释回退 | CX-2、DS-1 / models / M1、M3 | 支持/不支持的工具/流式能力可见;Run绑定实际模型;故障不串结果或重放工具 |
| C15 预算与用量 | 可设上下文/输出/步数,没有任务总成本控制 | 时间、token、费用、步骤、重试与worker总预算;无进展检测;可配置价格与未知usage标记 | CC-2、DS-2 / models、tasks / M1–M4 | 子任务/重试纳入总预算;耗尽时保存检查点;缺失usage不伪造准确费用 |
| C16 多代理协作 | 单会话串行执行,没有worker调度 | 独立问题分工、上下文/预算隔离、独立worktree、结果与证据汇总、冲突合并 | CC-1、CU-3、DS-2、WB-1 / tasks、workspaces / M4 | 两个独立任务并行完成可合并;重叠编辑被发现;父任务停止可传播 |
| C17 REA能力研究 | 产品内展示既有研究报告,没有REA执行入口 | 目标清单、版本/hash、inventory/analyze/trace、分析job、Evidence查看与研究结论登记 | 全部目标 / tools、artifacts / M3 | 从能力结论能追到确切工件与Evidence;静态/运行证据和未知范围分别显示 |
| C18 Purge增强实验 | 外部目录有1.1.78/1.1.81,未集成或验证收益 | 独立实验profile、兼容版本、策略分类、开关/回退和同题对照;有效机制按接口采纳 | PG-1 / engines、models、tasks / M3 | 同模型/任务/预算重复比较成功率、成本与错误修改;插件失效可恢复基线 |
| C19 外部代理引擎 | 尚无Codex/DeepSeek运行时适配 | EngineAdapter管理握手、会话/进程、事件、停止/恢复和真实能力限制 | CX-2、DS-1 / engines / M3 | 引擎事件有Task/Run映射;协议缺失能力明确;root idle不直接判工程完成 |
| C20 浏览器与界面验证 | 开发者用Playwright验证产品,代理尚无浏览器工具 | 可选浏览器工具、页面状态/截图、请求/控制台结果、交互验收与开发预览 | CU-2、WB-1 / tools、artifacts / M3–M6 | React真实交互/失败场景可复现;结果含页面与任务身份,不只保存截图 |
| C21 桌面产品 | 已有Windows浏览器便携包,没有原生桌面壳 | Electron Windows安装/升级、后台执行宿主、工程/终端/diff/成果区和任务通知 | 产品方案 / desktop、gateway / M5 | 关闭窗口后可再查看任务;安装/升级保留数据;宿主进程生命周期明确 |
| C22 手机伴随端 | 已验证390px响应式界面;仅绑定电脑127.0.0.1 | PWA任务/进度/审查入口;配对网关与设备撤销;共用Run与事件游标 | 产品方案 / mobile、gateway / M5 | 实际手机连电脑并审查同一任务;断线补发不重复执行;失效设备不能授权 |
| C23 数据、配置与发布迁移 | Key仅内存、JSON与备份;便携/源码包已验证 | 用户主动配置凭据、系统凭据存储、数据库迁移/回退、版本兼容和日志导出 | CX-2、DS-1 / storage、gateway、desktop / M1、M5–M6 | 升级副本与回退可验证;导出无Key;任务和来源不丢失 |
| C24 工程能力评测 | 基础行为/UI及少量真实模型场景 | 封存真实工程任务、独立验收、重复运行和错误分类;追踪人工介入/成本/恢复 | 全部目标 / evaluation、artifacts / M1–M6 | 发布能力可链接实际任务记录;报告含失败样本与环境,不能只展示成功演示 |
| C01 需求与任务规划 | Task固定目标/追加要求、Run身份和不可变依赖步骤/真实证明已交付;跨worker任务调度待做 | 固定目标、约束和完成标准;拆任务、判断依赖、接收追加要求与重规划 | CX-1、WB-1 / tasks / M1–M2 | 追加要求后仍保留原目标;任务完成由验收条件决定 |
| C02 执行状态与恢复 | SQLite事件/状态、原工件、Task检查点、unknown核对及显式同Task续接已交付;不重放已完成副作用 | 独立Task/Run/Step身份、持久事件、检查点与未知结果核对 | CC-2、CU-1、DS-1 / storage、tasks / M1 | 在审批、调用、结果与落盘等切点中断;恢复不重复已完成副作用 |
| C03 工程识别与索引 | 项目识别、JS/TS/React声明/导入索引与增量核对、Python候选已交付;完整语义/运行时依赖待做 | 识别语言、入口、规则、脚本、依赖和Git;增量索引;符号/引用与必要的语义检索 | CU-2、DS-3 / context / M2 | 对Node/React/Python样本定位实际修改入口;代码变动后索引更新 |
| C04 上下文与项目归属 | 有界原请求、工作依据/分页、实际工具反馈、来源记忆及Task/Run归属已交付 | 为Run组合需求、规则、代码、错误与验证结果;引用来源;严格绑定项目 | CC-2、CU-1 / context、contracts / M1–M2 | 切换两个项目/任务后,错误与工具结果仍归属正确;大输出用工件引用 |
| C05 长期项目记忆 | 项目持久记忆、来源/工程指纹、到期/归档、不完整扫描降级、脱敏请求快照与Task恢复包已交付 | 自动提取/归并、更长任务恢复与真实模型收益继续验收 | CC-2、WB-1 / context、storage / M4 | 长任务压缩/重启后完成标准不丢;过期代码事实被更新或标记 |
| C06 工具、Skills与MCP | 自有工程工具、固定REA、版本化本地Skills和项目所属受控stdio MCP已交付;远程MCP/任意插件安装待做 | 远程MCP、更多实际协议及插件安装/调用有来源和范围 | CC-1、CU-1、WB-2 / tools / M3 | 未知工具、错误参数、工具失联有原调用身份;新工具可按契约接入 |
| C07 执行控制 | 项目路径/原字节、逐次命令审批、预算、所属活动准入、停止和unknown门控已交付;OS沙箱与远程设备权限待做 | 操作范围、授权记录、预算、进程所有权、并发准入和取消传播;沿用用户已授权的范围 | CC-1、CU-1、DS-1 / tools、tasks / M1–M4 | 过期授权不能用于新Run;停止只处理所属任务;修改后的输入重新校验 |
| C08 精确编辑与多文件变更 | 精确文本patch、创建/修改、封存多文件变更集、逐文件恢复已交付;改名/删除及自动合并待做,跨文件非原子 | patch、创建/改名/删除、多文件ChangeSet、基线校验、隔离应用与可恢复合并 | CU-2 / workspaces / M2 | 跨4文件修改保留无关内容;外部改动引起冲突;部分失败有明确状态 |
| C09 用户工程Git | 用户工程Git预览/所属提交、私有索引、独立worktree及原修改保护已交付;自动合并/归档待做 | 仓库/dirty状态、基线、任务分支/worktree、diff、审查后提交与合并 | CX-1、CU-3、DS-3 / workspaces / M2 | 保留用户已有修改;每个交付可定位commit;冲突合并可审查 |
| C10 长命令与开发服务 | 终端60秒/有界输出;工程检查、REA/MCP/引擎/浏览器有所属job、原工件/退出与停止;后台开发服务待做 | 持久job、完整日志工件、心跳/退出码/超时、后台开发服务及所属进程树 | DS-3、CU-3 / tools、artifacts / M2 | 超过60秒的检查可观察和停止;失联结果标未知;重启不擅自再运行 |
| C11 真实检查与TestRunner | 实际Node/React/Python检查计划、测试数量/退出码/原日志、源码/范围绑定已交付;零测试/跳过不能算通过 | 识别真实项目检查;计划绑定planId,执行返回jobId;保存退出码、测试数和报告 | DS-3 / tools / M2 | 构建失败、零测试、拒绝/取消均不算通过;结果绑定相同工作区与ChangeSet |
| C12 修复与代码审查 | 实际失败反馈、有界修改/修复、无进展暂停、审查/Git门槛已交付;独立多模型复核与稳定自主完成率待验 | 从真实失败定位并有限轮修复;核对需求、diff、回归与风险;保留失败历史 | CX-1、DS-3 / tasks、workspaces / M2–M4 | 能复现的失败先出现再修复;实际重跑通过;审查不能用“模型说成功”代替 |
| C13 可交付成果 | Task所属diff/Git/检查/浏览器原工件与交付报告已交付;完整真实任务基准与专业交付指标待做 | 需求与结果对应、diff/commit、测试报告、运行方法、限制与工件引用 | WB-1 / artifacts / M2–M6 | 用户按交付说明能启动并验证;报告与实际文件、检查及commit一致 |
| C14 模型协议与路由 | 多连接兼容Chat协议池、Run固定路由、轮询/权重/租约/冷却、已验证格式导入已交付;stream:false,多协议/多模态待做 | ModelProvider能力探测、流式文本/工具参数、模型配置快照、按任务角色路由与可解释回退 | CX-2、DS-1 / models / M1、M3 | 支持/不支持的工具/流式能力可见;Run绑定实际模型;故障不串结果或重放工具 |
| C15 预算与用量 | 完整请求预检、Run/Task累计步数/token预算、重试/用量账本及unknown usage已交付;费用价格与worker总预算待做 | 时间、token、费用、步骤、重试与worker总预算;无进展检测;可配置价格与未知usage标记 | CC-2、DS-2 / models、tasks / M1–M4 | 子任务/重试纳入总预算;耗尽时保存检查点;缺失usage不伪造准确费用 |
| C16 多代理协作 | 单会话串行工具执行,已有独立worktree;没有worker调度、成果合并与父子停止传播 | 独立问题分工、上下文/预算隔离、独立worktree、结果与证据汇总、冲突合并 | CC-1、CU-3、DS-2、WB-1 / tasks、workspaces / M4 | 两个独立任务并行完成可合并;重叠编辑被发现;父任务停止可传播 |
| C17 REA能力研究 | 固定REA6.8.0的ASAR/PE三项静态操作、所属job/目标SHA256/原Evidence已交付;其余工具与运行trace待做 | 目标清单、版本/hash、inventory/analyze/trace、分析job、Evidence查看与研究结论登记 | 全部目标 / tools、artifacts / M3 | 从能力结论能追到确切工件与Evidence;静态/运行证据和未知范围分别显示 |
| C18 Purge增强实验 | DSH-purge来源/机制研究已登记,未集成工程执行或验证收益;兼容组合与同题对照待做 | 独立实验profile、兼容版本、策略分类、开关/回退和同题对照;有效机制按接口采纳 | PG-1 / engines、models、tasks / M3 | 同模型/任务/预算重复比较成功率、成本与错误修改;插件失效可恢复基线 |
| C19 外部代理引擎 | 固定DeepSeek0.2.1-rc.2的六方法连接/所属进程/原协议已交付;Codex适配及引擎prompt/工程写入待做 | EngineAdapter管理握手、会话/进程、事件、停止/恢复和真实能力限制 | CX-2、DS-1 / engines / M3 | 引擎事件有Task/Run映射;协议缺失能力明确;root idle不直接判工程完成 |
| C20 浏览器与界面验证 | 代理固定静态构建、真实Edge交互/DOM/PNG/原报告与Task绑定已交付;任意在线浏览与电脑操作待做 | 可选浏览器工具、页面状态/截图、请求/控制台结果、交互验收与开发预览 | CU-2、WB-1 / tools、artifacts / M3–M6 | React真实交互/失败场景可复现;结果含页面与任务身份,不只保存截图 |
| C21 桌面产品 | Windows浏览器便携/源码包及保留原进程升级已交付;Electron安装版、后台宿主/窗口拆分待做 | Electron Windows安装/升级、后台执行宿主、工程/终端/diff/成果区和任务通知 | 产品方案 / desktop、gateway / M5 | 关闭窗口后可再查看任务;安装/升级保留数据;宿主进程生命周期明确 |
| C22 手机伴随端 | 390px响应式界面已验证;电脑仅绑定127.0.0.1,真实手机PWA配对/撤销待做 | PWA任务/进度/审查入口;配对网关与设备撤销;共用Run与事件游标 | 产品方案 / mobile、gateway / M5 | 实际手机连电脑并审查同一任务;断线补发不重复执行;失效设备不能授权 |
| C23 数据、配置与发布迁移 | Key仅内存、SQLite/旧JSON迁移、不可变包/保留Provider与历史升级已交付;本机加密凭据、完整回退和设备配置待做 | 用户主动配置凭据、系统凭据存储、数据库迁移/回退、版本兼容和日志导出 | CX-2、DS-1 / storage、gateway、desktop / M1、M5–M6 | 升级副本与回退可验证;导出无Key;任务和来源不丢失 |
| C24 工程能力评测 | 行为/UI与Node/React/Python实际样本、失败/人工介入证据已保留;封存成套基准、重复评测与整体稳定完成率待做 | 封存真实工程任务、独立验收、重复运行和错误分类;追踪人工介入/成本/恢复 | 全部目标 / evaluation、artifacts / M1–M6 | 发布能力可链接实际任务记录;报告含失败样本与环境,不能只展示成功演示 |
### 1.6.1 M2实际工作依据切片
+78
View File
@@ -0,0 +1,78 @@
# 里程碑回顾与修复审查
日期:2026-10-30。审查基线为1.12.0实现3f6797b和交付记录ca9042e;本地正式服务为已验收的不可变发行20261030124454517。用户要求本阶段先审查、修复及记录计划,随后再开发新能力。
本报告区分历史发行证据、当前源码检查、当前隔离复验及未完成项。各版本原报告和失败记录保持原字节,不把过去的局部验证推断成整体自主交付能力。
## 执行与产物
- 后续计划:逐项记录范围、前置条件、真实完成标准及未知能力。
- 里程碑:核对计划、模块、原验收与实际部署,整理阶段遗留项。
- 可靠性与安全:检查本地HTTP权限、项目文件/Git范围、任务/审批归属、持久恢复、模型池/凭据、Skills/MCP/REA/外部引擎和交付证据。
- 修复:先以独立端口、临时项目和自有凭据夹具复现,再做相关验证;不读取真实Key、不发送真实模型请求,不普通重启正式服务。
- 收尾:本地小步提交、必要的补丁发行、最终证据与未完成能力说明。
## 当前进度
已核对1.12.0交付:357/357独立源码、150组完整界面和部署后8个页面通过;原PID19084/Provider以及15项目、26会话、2505记录、4476事件在升级时保持。此记录是历史发行基线,本阶段新的检查另行记录。
后续顺序、前置条件与完成标准已保存于[NEXT-STEPS.md](NEXT-STEPS.md)。80项相关专项、额外实际配置HTTP故障检查、旧发行进程升级/Edge8组已通过。首个源码候选368/369,唯一失败来自Windows尾点提案提前拒绝后的旧测试假设;追加硬链接复现、补强及路径/提案/变更集29项复验通过。包含全部修复的新构建20类150组UI全部通过、0页面/请求异常;最终独立源码和正式补丁部署另行登记,首失败候选不得部署。
## 逐版本对照
下表数字来自各版本原报告,不是本轮重新运行所有旧版本。原真实模型失败保持失败;各版本的源码、便携、正式部署及原失败详见对应QA。
| 节点 | 原计划与已实现能力 | 原验收 | 原真实工程及遗留项 |
| --- | --- | --- | --- |
| [1.0](QA-1.0.md) | 基础对话、工具、代码提案/审批、诊断及旧会话 | 34项测试、19组UI | 是基础可运行版本,尚无完整工程交付与可靠Task内核 |
| [1.1](QA-1.1.md) | Task/Run/Tool、SQLite/工件、预算、实际检查及中断恢复 | 68项、26组UI | 内核与检查链路交付;真实工程完成率需另验 |
| [1.2](QA-1.2.md) | 当前Task的Git预览/确认、所属提交、交付报告与unknown核对 | 92项、35组UI | 真实Node样本四文件修改/检查/Git通过;三类工程完整基准未完成 |
| [1.3](QA-1.3.md) | 工程声明/导入索引、精确补丁、保留原格式与Git范围 | 112项、43组UI | 真实精确四文件样本通过;索引不代表完整运行时依赖,成本收益未对照 |
| [1.4](QA-1.4.md) | 封存多文件变更集、整批预检、逐文件账本与显式回退 | 141项、54组UI,13个实际退出切点 | 真实样本工程交付通过,模型最终因token预算暂停;跨文件写入并非原子事务 |
| [1.5](QA-1.5.md) | 完整请求预算、有界上下文、原工件与历史保留 | 158项、62组UI | 原真实模型27次响应后超时,无提案;不能记为自主交付 |
| [1.5.1](QA-1.5.1.md) | 按实际预算分页、有界最新结果、重复读取暂停 | 163项、63组UI | 真实代码/检查通过,Git预览后循环;人工核对收尾,不算全自主 |
| [1.6](QA-1.6.md) | 独立Git工作区、身份/创建unknown、保留已有修改及审查收尾 | 192项独立源码、72组UI | 工作区执行链路通过,原真实模型仍无提案;自动合并/归档未交付 |
| [1.6.1](QA-1.6.1.md) | 原结果可核对的Task工作依据与预算观察 | 200项独立源码、74组UI | 界面和观察通过,真实工作依据仍被挤出;下一补丁修复 |
| [1.6.2](QA-1.6.2.md) | 工作依据优先保留、旧发送视图兼容和原工件分页 | 205项独立源码、74组UI | 原依据保留已修;真实模型仍重复查询且无提案,原失败未改写 |
| [1.7](QA-1.7.md) | 最新工具反馈/无进展暂停、Python结构化检查、React工程验证 | 224项独立源码、87组UI | 受控React/Python执行样本通过;原真实模型no_progress失败保留 |
| [1.8](QA-1.8.md) | 固定REA 6.8.0 stdio与ASAR/PE三项静态操作、所属原Evidence | 244项独立源码、98组UI、真实REA三操作 | 真实Node四模块工程Task的14项检查/build/Git通过;REA其他工具与引擎执行未交付 |
| [1.9](QA-1.9.md) | 固定DeepSeek Harness 0.2.1-rc.2 SDK连接、所属进程/原协议/停止 | 261项独立源码、108组UI、六方法连接核验 | 真实React/Python样本通过;尚未向引擎发送prompt/开放工程写入,Purge未做收益对照 |
| [1.10/1.10.1](QA-1.10.md) | 封存静态构建与真实Edge交互/PNG;热升级旧工件对象API修复 | 274/276项独立源码、118组UI | 实模桌面8/8,390px选择器匹配三个按钮而失败;整体真实页面任务未完成 |
| [1.11](QA-1.11.md) | Task不可变依赖计划、所选范围真实证明、唯一DOM定位/字段读取 | 298项独立源码、126组UI | 源码/便携/正式部署通过;真实模型续接按用户选择暂缓 |
| [1.12](QA-1.12.md) | 模型中心/池、已验证格式导入与可关闭同步、巡检、来源记忆、检查点、Skills、stdio MCP | 357项独立源码、150组UI、部署后8个页面 | 正式部署通过;实模工程暂缓;WorkBuddy未知格式、加密凭据、跨协议/流式/多模态待做 |
## 总计划实现情况
| 计划 | 本轮核对后的状态 | 下一实际门槛 |
| --- | --- | --- |
| M0工程基线 | 已建立本地Git、原发行、来源/能力矩阵和研究规范 | 后续独立实现通过相关检查后及时本地提交 |
| M1可靠内核 | Task/Run、SQLite、审批/预算、原工件与恢复已交付;本轮补持久化/升级/凭据传递边界 | 保持未知副作用核对、失败回滚和原进程升级回归 |
| M2真实工程闭环 | 代码、检查、浏览器、Git和报告已接通;部分真实样本成功、部分失败 | 固定Node/React/Python各两个任务,独立验收并统计人工介入/完成率,不能用测试数量代替 |
| M3研究与引擎 | REA三项与DeepSeek连接已交付;受控引擎工程执行和Purge收益未完成 | 固定实际写入/取消/审批协议,隔离工程与同题对照 |
| M4长任务/多worker | 记忆、上下文工件、检查点和独立worktree已交付;worker调度/成果合并未完成 | 两个受限worker及真实冲突/停止/恢复验收 |
| M5桌面/手机 | Web界面及390px布局已验证;安装版与真实设备配对未实现 | 单实例后台宿主、保留数据更新、可撤销配对和断线补发 |
| M6专业交付 | 不可变包、输入/构建/日志散列、独立源码/便携验证已有 | 加密凭据/隔离、真实任务质量指标、候选回滚及隔离自我迭代 |
## 本轮确定问题与修复
| 编号 | 问题与复现 | 修复范围 | 当前验收 |
| --- | --- | --- | --- |
| A01 | 项目内junction指向.git或credentials可绕过词法保护,创建预检也允许;受保护文件硬链接为AGENTS.md可被普通读取及自动规则加载 | safePath同时核对真实目标/父目录、凭据目录和Windows尾点/空格别名;常规文件要求单一路径,清单跳过多硬链接;普通项目内链接及独立副本可用 | 原junction/硬链接复现均保留;新增读取/提案/清单/自动规则与旧提案兼容检查,相关29/29通过,最终回归另验 |
| A02 | SQLite保存失败后模型注册内存状态与已提交数据库不一致;选择/运行模式及批量导入可能部分生效 | 保存失败恢复已提交注册;模式与选择同事务;整批连接/来源/凭据引用回滚;成功提交后才替换原Key | 原复现milestone-model-save-initial-repro.json;10项模型中心通过,最终回归另验 |
| A03 | 热升级保留已初始化ModelHub/Importer对象却不更新方法;模型探针未参与活动门控 | 保留对象/凭据/预览并更新原型,失败恢复旧原型;在途请求/队列/租约阻止配置变更与升级 | 原对象/失败回滚和探针专项通过;实际旧1.12初始化对象升级及原凭据/预览保持通过 |
| A04 | 本地审批页面可被外部页面嵌入,缺少同源脚本/连接策略 | 禁止frame嵌入、仅同源脚本/API、禁止object/base、no-referrer;已有动态布局保留内联style | HTTP与实际Edge旧/新iframe对照、外部脚本拒绝、1280/390正常页面及PNG Blob通过;完整UI150组通过 |
| A05 | PowerShell、npm及Python检查继承未知宿主凭据环境变量 | 统一仅传系统/运行路径和内核显式字段,保留Python检查隔离设置 | 三条实际执行链路原标记复现及新增shell/npm/Python实际检查通过,Node/Python相关回归通过 |
初始复现与失败记录都在独立`.qa/milestone-*`目录;不涉及用户项目、真实Key或模型请求。目录和报告将在最终身份清单中逐项绑定。
## 审查覆盖与结论边界
- HTTP:实际本地Host/Origin/token、JSON限制、事件游标/慢客户端、只读与副作用路由;新增防嵌入和同源策略。当前本机token不等同于OS沙箱或远程设备权限。
- 文件/Git:词法与实际路径、受保护目录、提案原字节、ChangeSet预检/恢复、私有Git索引/所属提交、工作区HEAD/身份与unknown。保留已有用户修改;外部并发替换链接的完整OS隔离仍属后续能力。
- 执行:Task/Run/审批归属、稳定工具身份、请求预算、检查计划/测试数量与原日志、进程取消和未知结果;补齐宿主环境隔离。当前审批命令仍使用宿主账户权限。
- 计划/上下文/恢复:有界原请求、最新工具反馈、来源记忆、Skills目录/原修订、检查点及证明原工件;来源/要求/源码改变导致失效,不重写历史发送证据。
- 模型:配置事务、选择/Run固定快照、队列/租约/冷却、401/429、停止、原请求账本与导入预览指纹。Key目前仅在内存,实际多协议/真实模型工程另验。
- REA/MCP/引擎/浏览器:固定身份与协议、所属进程、受控范围、原报告/PNG/退出、unknown核对和不重放。连接核验、静态构建和夹具不宣称为任意引擎/桌面操作。
- 依赖:2026-10-30 npm audit的153项依赖元数据中0已知漏洞;原JSON与SHA256单独保存。此结果不证明不存在未知漏洞;REA等随包依赖另按锁/清单和实际运行验证。
当前确定修复、最终全量回归、实际UI与补丁部署需全部完成后才能关闭本轮审查。后续新能力按NEXT-STEPS执行,真实模型验收继续暂缓。
+51
View File
@@ -0,0 +1,51 @@
# 后续执行计划
更新:2026-10-30。当前交付基线为1.12.0;本轮先完成[里程碑审查](MILESTONE-AUDIT.md)及补丁验收,再进入下一阶段。历史研究、局部测试与连接核验不能代替真实工程交付。其他项目不进入本计划。
产品目标是让用户给出需求后,工作台能保留项目基线,完成定位、计划、修改、检查、必要修复、审查和Git交付,并在停止、故障、断线后明确保留已完成结果。桌面与手机共享同一执行宿主、任务和证据。
## 执行顺序与完成门槛
| 顺序 | 开发范围 | 前置条件 | 完成标准 |
| --- | --- | --- | --- |
| 0:审查闭环 | 对照1.0–1.12计划、模块、原始报告;修复确认问题;补丁发行和回滚验证 | 保留旧发行、历史失败和正式服务连接 | 当前源码、实际进程、桌面/390px、独立源码及便携包通过;正式升级保留原Provider及全部历史;未完成能力明确登记 |
| 1:隔离执行与多worker | 父任务拆分、所属worker、独立worktree、预算分配、停止传播、结果归集及冲突审查 | 当前Task/Run、工作区、计划和未知结果恢复契约 | 两个worker各自在隔离工程完成真实修改/检查;重叠修改被识别;父任务停止后无所属进程继续写入;超时/崩溃不重放工具;只合并已审查成果 |
| 2:引擎执行与增强对照 | 固定版本的Codex/DeepSeek适配、受控工具桥;Claude Code/Cursor/WorkBuddy机制研究;DSH-purge实验 | 外部SDK的实际审批、取消、写入与事件契约;阶段1隔离工作区 | 引擎能完成一个指定工程任务,原输入、工具与结果可追溯;Purge按同模型、同题、同预算进行有/无插件对照,记录失败和人工介入;仅采纳有证据的机制 |
| 3:Windows桌面 | 单实例、后台宿主、项目/终端/审查入口、安装包与保留数据更新 | 宿主与客户端生命周期拆分;本机凭据加密;稳定迁移契约 | 关闭/重开窗口不丢任务;重复启动复用宿主;安装/升级/卸载的数据行为可核验;崩溃后显式恢复;旧版本可回退 |
| 4:手机伴随端 | PWA、一次性配对、设备身份、撤销、同Run观察/审批、断线游标补发 | 设备认证、加密传输、审批身份/幂等、速率限制;阶段3宿主 | 在真实手机完成配对和撤销;断线后按seq恢复;旧Run/过期/重复审批拒绝;手机和桌面显示同一结果;手机不直接执行宿主文件与命令 |
| 5:模型与多模态 | 更多实际协议、流式取消、按能力选路、可恢复池策略、视觉/语音/附件、项目缓存 | 每种协议的工具调用/用量/取消实测;凭据和附件隔离 | 图像、语音、文本各有真实样本;流式断开保留unknown而不重复副作用;不兼容成员不被调度;缓存绑定项目/权限/源码版本并可清除;429/401/超时/冷却有独立证据 |
| 6:自我迭代与专业交付 | 隔离分支生成候选、固定评测集、多模型复核、质量/成本对照、升级候选和回滚 | 可复现工程基准、完整质量门槛、阶段1隔离和阶段3更新 | 工作台可提出并验证自身改进;候选不直接改运行中的发行;回归或独立验收失败不能发布;保存原差异、指标和回退点;采纳须经用户已有权限规则 |
阶段间可并行完成只读研究和文档,涉及项目写入、权限、发行的实现按依赖推进。每个切片提交真实功能、失败分类、验收证据和本地Git提交;不自动推送远端。
## 下一开发切片:隔离worker
先实现两个受限worker的完整流程,再扩展并发规模。父Task固定目标、验收条件和总预算;子任务保存范围、工作区、所属进程与原证据。调度器负责分配/释放预算和租约,不让模型自行绕过权限。
1. 建立父子任务与worker状态契约:queued、running、stopping、completed、failed、unknown;完成状态需关联真实工件与检查结果。
2. 在独立worktree创建工程快照,保留用户原暂存/工作区修改;来源身份改变、创建结果未知和未审查冲突阻止下一次写入。
3. 实现预算、并发与停止传播;父任务取消后结束所属进程树,等待退出并保存原结果,失联状态显式核对。
4. 归集diff/测试/报告,校验来源与基线;分离修改可审查合并,重叠修改显示具体文件与冲突,不覆盖父工程。
5. 使用Node、React、Python各一个固定工程任务,验证正常、测试失败后修复、冲突、停止、宿主退出和恢复。真实模型语义验收待用户在工作台配置后单独执行。
## 跨阶段必须保持的能力
| 能力 | 当前基础与后续要求 |
| --- | --- |
| 凭据 | 当前Key仅在内存;后续采用Windows本机加密或系统凭据库,导入由明确格式和用户授权驱动,不读取未知登录态;重启/密钥失效可见 |
| 权限与执行范围 | 当前本地token、同源请求、审批和项目路径保护;增加任务/设备/工具作用域及OS隔离;模型、插件与外部引擎均受执行内核控制 |
| 恢复与可靠性 | 保留Task/Run、工件、计划、检查点和unknown;增加故障注入、迁移/回滚演练、所属进程核对及资源上限,不自动补造结果 |
| Skills/MCP/插件 | 保留来源、版本、启停、范围与原调用证据;实际协议逐项接入,远程服务/安装和未知副作用独立核验 |
| 项目记忆/上下文/缓存 | 当前来源复核记忆和有界请求工件;后续缓存按源码/模型/权限版本失效,长任务依据与原结果始终可追溯 |
| 浏览器/电脑操作 | 当前固定静态构建与真实Edge验收;后续真实浏览器/桌面操作限定目标、登录态和权限,并验证停止与敏感操作审批 |
| 号池 | 当前兼容Chat协议、轮询/权重/租约/冷却、已验证格式导入和可关闭同步;后续按能力、上下文、成本选路,跨协议实际验证,不承诺上游永不中断 |
| 巡检 | 当前只读运行巡检;增加任务账本、工件、来源、池健康、空间/进程/迁移检查。修复与巡检分别授权,巡检不偷偷执行模型或项目命令 |
| 交付质量 | 固定任务基线、需求、原测试/构建/交互、实际diff/Git和报告;用完成率、人工介入、失败类型、耗时及已报告用量评价改善 |
## 研究依据与未知项
REA用于发行工件或运行证据,完整公开源码直接分析。研究以实际版本、来源、问题、证据和拟落地模块登记,见[研究路线](RESEARCH-ROADMAP.md)。参考产品的名称不证明对应能力已经接入。
DSH-purge源码位于`D:\Ai-project\REA逆向\dsh-purge-1.1.78-windows`;DeepSeek工程位于`D:\Ai-project\DeepSeek-Harness-Desktop`。它们与ModelProvider分层管理。当前DeepSeek六个SDK方法只完成连接核验,Purge的工程收益尚未获得对照证据。
尚未交付:OS沙箱、任意外部引擎工程写入、未知WorkBuddy格式自动导入、本机加密持久凭据、多协议流式/多模态、Windows安装版、真实手机配对,以及稳定的全自主工程成功率。按上表逐项实现和验收,不用界面入口或夹具结果替代完成。
+2 -2
View File
@@ -1,6 +1,6 @@
# Agent Workbench 后续产品与执行开发方案
日期:2026-10-30。本文保留完整产品规划;1.8.0已交付内置REA与真实Node工程修复,1.9.0已完成DeepSeek Harness连接核验、独立发行和真实React/Python工程样本。1.10.1完成276项独立源码、不可变发行和保留原Provider升级,实模页面仍保留手机定位失败。1.11.0已完成持久依赖计划、原证据核验、实际DOM唯一定位与页面读取进展,最终独立源码298/298、126组界面、源码/便携与正式本地升级通过;真实模型续接按用户选择暂缓。下一阶段的专业UI、连接注册、模型池、导入和巡检详见[1.12执行方案](PLAN-1.12.md)。实际实现、验收与部署分别见[能力矩阵](CAPABILITIES.md)及[验证记录](VALIDATION.md),完整阶段目标继续实施。
日期:2026-10-30。本文保留完整产品规划;1.8.0已交付内置REA与真实Node工程修复,1.9.0已完成DeepSeek Harness连接核验、独立发行和真实React/Python工程样本。1.10.1完成276项独立源码、不可变发行和保留原Provider升级,实模页面仍保留手机定位失败。1.11.0已交付持久依赖计划、原证据核验与实际DOM唯一定位;1.12.0已正式交付模型中心/池、已验证格式导入、巡检、项目记忆、检查点、Skills和受控stdio MCP,357/357独立源码及150组界面通过。真实模型续接按用户选择暂缓。当前先做[里程碑审查](MILESTONE-AUDIT.md),再按[后续执行计划](NEXT-STEPS.md)推进隔离worker、受控引擎、桌面/手机及其余能力。实际实现、验收与部署分别见[能力矩阵](CAPABILITIES.md)及[验证记录](VALIDATION.md)。
## 1. 产品目标
@@ -29,7 +29,7 @@ Purge的源码机制不证明模型权重或工程成功率已经提升。增强
现有基础:React工作台、兼容模型连接、Task/Run/Tool及预算、SQLite/工件恢复、工程索引与精确补丁、npm实际检查、单文件/整批审核、Git和报告。1.4.0增加原/新字节快照和逐文件执行账本、整批预检及显式恢复;各版本实测证据单独登记。
正式基础增加了1.11的Task所属依赖计划及原结果核验。1.12源码已增加多连接与兼容Chat协议模型池、本地格式导入/巡检,以及带来源核对和有界请求快照的[项目记忆](PROJECT-MEMORY.md),尚未作为新不可变发行部署。多worker、动态REA/MCP/外部引擎注册、桌面壳和手机配对继续开发;已有不同阶段的Node/React/Python工程样本,整体自主交付能力按真实任务评估。1.7.0已接入Python专用检查和混合范围选择,并完成受控React/TypeScript实际类型/构建/交互/Git样本;来源和验收分别见[Python契约](PYTHON-CHECKS.md)、[React基准](../benchmarks/react-delivery-v1/README.md)及[QA-1.7](QA-1.7.md)。1.6.0已有独立Git工作区创建/身份/恢复,自动合并和归档继续开发。批次支持文本创建/修改,不提供跨文件原子事务。正式1.11仍为原单连接,协议仍使用stream:false;跨协议与真实模型池工程验收另行记录。
正式1.12基础包括Task所属依赖计划与原结果核验、多连接兼容Chat协议模型池、本地格式导入/巡检、带来源核对和有界请求快照的[项目记忆](PROJECT-MEMORY.md)、Task检查点、本地Skills和受控stdio MCP。多worker、动态REA/外部引擎注册、桌面壳和手机配对仍待开发;已有不同阶段的Node/React/Python工程样本,整体自主交付能力按真实任务评估。1.7.0已接入Python专用检查和混合范围选择,并完成受控React/TypeScript实际类型/构建/交互/Git样本;来源和验收分别见[Python契约](PYTHON-CHECKS.md)、[React基准](../benchmarks/react-delivery-v1/README.md)及[QA-1.7](QA-1.7.md)。1.6.0已有独立Git工作区创建/身份/恢复,自动合并和归档继续开发。批次支持文本创建/修改,不提供跨文件原子事务。当前模型协议仍使用stream:false;跨协议与真实模型池工程验收另行记录。
工作台源码已建立本地Git,基线提交为`bac87da`,标签`v1.0.0-baseline`,继续按可审查小步提交。用户工程已有内置审查提交/证据报告;1.6.0补齐独立工作区,自动合并、归档及完整工程验收继续实施。
+36
View File
@@ -0,0 +1,36 @@
# 1.12.1 里程碑审查补丁验收
日期:2026-10-30。基线1.12.0/3f6797b/ca9042e保持原报告。修复提交:c1e9575(实际路径保护)、f4bd0ad(配置/升级、环境与HTTP边界、独立发行入口)、7b1f529(硬链接保护及旧Windows提案兼容验证)。
本补丁完成[逐里程碑审查](MILESTONE-AUDIT.md)和[后续执行计划](NEXT-STEPS.md),修复五类确认的安全与可靠性问题。当前完成专项与完整UI,独立源码、最终不可变包和正式升级仍按下述流程验收;完成前不登记为正式交付。
| 验证 | 当前结果 | 原证据 |
| --- | --- | --- |
| 路径/Git及凭据内部junction保护 | 实际旧逻辑复现,修复后相关检查通过 | [原复现](milestone-audit-path-initial-repro.json) |
| 硬链接及路径/变更集追加复核 | 自有受保护规则硬链接实际复现;修复后读取/提案/清单/自动规则及旧提案兼容29/29、0失败/取消/跳过 | [硬链接原复现](milestone-audit-hardlink-initial-repro.json)、[追加复验](milestone-audit-path-followup-validation.json)、[原TAP](milestone-audit-path-followup-tests.tap) |
| 模型SQLite提交失败、选择/模式及批量导入回滚 | 实际旧逻辑复现、10项模型中心和额外HTTP故障/非法Key/重试通过 | [原复现](milestone-audit-model-save-initial-repro.json) |
| 终端/npm/Python环境边界 | 三条旧执行链路都继承3个自有标记;新实际执行专项通过 | [原复现](milestone-audit-environment-initial-repro.json) |
| 相关源码/API/实际退出与升级 | 80/80、0失败/取消/跳过 | [专项](milestone-audit-targeted-tests-validation.json)、[原TAP](milestone-audit-targeted-tests.tap) |
| 真实旧1.12进程→补丁、已初始化Hub/Importer、CSP与Edge | 8组通过;原PID/预览/两类夹具凭据保持;失败批量不部分导入;iframe/外部脚本拒绝;1280/390与PNG Blob可用 | [实际进程与Edge](milestone-audit-security-ui-validation.json)、[验收器](milestone-audit-security-ui-verifier.mjs) |
| 全部功能界面 | 包含硬链接追加修复的新构建20类150组全部通过、0页面/请求异常;原首轮UI另存,不覆盖失败候选依据 | [全部UI](milestone-audit-all-ui-validation.json)、[首轮基线](milestone-audit-first-candidate-baseline.json) |
| 首个独立源码候选20261030150404738 | 离线构建/6份构建匹配与REA/DeepSeek/Edge通过;368/369,旧测试假设提前拒绝时机导致唯一失败;此候选不部署 | [首轮失败](milestone-audit-first-source-package-failure.json)、[首轮原TAP](milestone-audit-first-source-package-tests.tap) |
| npm依赖 | 153项元数据、0已知漏洞;lock仅根版本变化,依赖内容与原审计严格相同 | [原审计](milestone-audit-npm-audit.json)、[摘要](milestone-audit-dependency-summary.json) |
专项使用Node24.22.1、独立Python3.10.11/pytest8.4.2、自有项目/本地HTTP凭据和独立端口。实际Edge验证的两张界面/拒绝嵌入截图已目视核对,功能UI包含真实浏览器结果/PNG展示;没有使用正式模型连接。
## 最终发行流程
1. 冻结当前已提交源码、测试、脚本和构建;从不可变源码ZIP独立离线安装/构建及完整回归,核对输入/6份构建/逐阶段日志。
2. 封存原结果后仅更新交付文档;最终不可变ZIP的实现输入、原生依赖、构建必须与通过的独立源码相同。REA-RUNTIME只允许明确登记的builtAt打包时间变化。
3. 最终归档CRC/清单/目录/散列、实际便携Node/npm/REA/DeepSeek/Edge验证后,核对正式服务当前PID、公开配置和全部项目/会话/记录/事件及每份导出。当前只读实测1.12.0/PID23172、hasKey=true、16项目/20可见会话,无忙任务,见[当前服务身份](milestone-audit-service-preflight.json);旧PID19084是历史交付记录,最新全部数据需重新快照。
4. 保留原Provider进行升级;部署后桌面/390px只读UI与数据库复核,关闭9229,保存最终身份和本地Git提交。
## 原失败与范围
首个源码候选的Windows尾点路径已在当前提案入口拒绝,旧测试仍试图先生成提案后核对批次错误,导致368/369。已分别核对当前入口提前拒绝和旧宿主保留提案在批次层拒绝,相关29项通过;新增硬链接补强不在首候选中,须重新构建封包及独立完整回归。
首轮19/20的探针专项在成功探针之后提交不完整/config夹具而失败,修正完整地址/模型后80项专项通过。首轮Edge导入故障夹具用无效URL,来源解析按契约预先跳过,原失败保留[初始记录](milestone-audit-security-ui-initial-failure.json);改用超长自有凭据后实际回滚通过。
首次构建子进程PATH仍找到Node16,前置本次Node24后通过;三个验收入口原来未实现--help,曾额外运行一次旧版独立验收,未改历史docs或正式服务,现已增加先行帮助/未知参数拒绝。所有环境和工具查询失败见findings/progress,不用新结果覆盖旧记录。
真实模型请求为0。真实模型工程验收按用户选择暂缓。OS沙箱、加密持久凭据、任意引擎工程写入、Purge收益、未知WorkBuddy格式、多协议/流式/多模态、安装版/真实手机与稳定全自主完成率仍按NEXT-STEPS开发或验证。
+1 -1
View File
@@ -1,6 +1,6 @@
# REA与代理机制研究路线
日期:2026-10-29。本文交付的是研究与落地路线;当前没有完成五款客户端全部能力的逆向,也没有把Purge或外部SDK集成到工作台。
更新:2026-10-30。本文记录研究与落地路线;1.8已将REA三项静态分析接入工作台,1.9已将固定DeepSeek Harness六个SDK方法接入连接核验。五款客户端完整机制、受控外部引擎工程执行及Purge增强收益仍需研究和实测。当前先完成[里程碑审查](MILESTONE-AUDIT.md),后续按[NEXT-STEPS](NEXT-STEPS.md)推进。
来源身份登记见 [research/targets.json](../research/targets.json)。产品目标见 [PRODUCT-PLAN.md](PRODUCT-PLAN.md),能力ID见 [CAPABILITIES.md](CAPABILITIES.md)。
+12
View File
@@ -12,6 +12,18 @@
1.10.0完成274/274独立源码、13类118组界面、170份冻结输入及不可变源码/便携验证。实模首轮发现旧SQLite工件对象缺少新增字节读取API,原失败保留;1.10.1修复后最终独立源码276/276、174份冻结输入、实际随包REA/DeepSeek/Edge与便携/归档均通过。正式4318保留PID23080/Provider、14项目和全部24会话升级1.10.1。实模复验桌面8/8断言通过,390px因按钮选择器匹配三个元素而失败,210秒内未完成整体任务;18次响应/152837tokens及人工停止按原结果封存,见[QA-1.10](QA-1.10.md)。
## 1.9.0 · DeepSeek连接核验与真实工程样本
261/261独立源码、108组界面、不可变包与保留原Provider升级通过;固定DeepSeek Harness 0.2.1-rc.2的六个SDK方法与所属进程退出验证完成。真实React/Python工程样本分别保留实际Task、检查和Git证据。当前SDK只完成连接核验,未发送引擎prompt、开放工程写入或证明Purge提升,见[QA-1.9](QA-1.9.md)。
## 1.8.0 · 内置REA与工程修复
244/244独立源码、98组界面、源码/便携及正式升级通过。固定REA 6.8.0的ASAR/PE三项静态操作有实际Evidence与所属Task/Run/目标SHA256,真实Node四模块工程Task的14项测试、build与Git交付通过。其余REA工具和外部引擎完整执行不计入此版本,见[QA-1.8](QA-1.8.md)。
## 1.7.0 · 最新工具反馈与Python/React检查
224/224独立源码、87组界面及不可变发行通过;最新工具结果反馈、相同工程状态暂停、Python结构化测试/编译与检查范围选择已交付,受控React/TypeScript交互/检查/Git样本通过。原真实模型因no_progress暂停且无提案,不能登记为真实自主交付,见[QA-1.7](QA-1.7.md)。
## 1.6.2 · 工作依据保留与旧发送视图兼容
不可变发行20261029180412072的独立源码205/205项、离线安装和6份构建比对通过;74组界面及便携/ZIP验收通过。原PID23080、Provider、10项目/12会话和全部执行证据保留升级,模型配置与预算未改变。见[QA-1.6.2](QA-1.6.2.md)、[源码](working-retention-source-package-validation.json)和[升级](working-retention-upgrade-live-validation.json)。七份真实冻结请求全部保留四模块/README,原结果可核验;离线回放不算实际模型完成。
+169
View File
@@ -0,0 +1,169 @@
{
"success": true,
"version": "1.12.1",
"at": "2026-10-30T15:37:24.953Z",
"groups": 150,
"results": [
{
"script": "verify-task-plans-ui.mjs",
"success": true,
"groups": 8,
"durationMs": 9081,
"evidence": "docs/milestone-audit-task-ui-validation.json",
"log": ".qa/milestone-audit-verify-task-plans-ui.log"
},
{
"script": "verify-browser-ui.mjs",
"success": true,
"groups": 10,
"durationMs": 19968,
"evidence": "docs/milestone-audit-browser-ui-validation.json",
"log": ".qa/milestone-audit-verify-browser-ui.log"
},
{
"script": "verify-ui.mjs",
"success": true,
"groups": 10,
"durationMs": 19171,
"evidence": "docs/milestone-audit-basic-ui-validation.json",
"log": ".qa/milestone-audit-verify-ui.log"
},
{
"script": "verify-diagnostics-ui.mjs",
"success": true,
"groups": 9,
"durationMs": 19578,
"evidence": "docs/milestone-audit-diagnostics-ui-validation.json",
"log": ".qa/milestone-audit-verify-diagnostics-ui.log"
},
{
"script": "verify-execution-ui.mjs",
"success": true,
"groups": 7,
"durationMs": 24935,
"evidence": "docs/milestone-audit-execution-ui-validation.json",
"log": ".qa/milestone-audit-verify-execution-ui.log"
},
{
"script": "verify-delivery-ui.mjs",
"success": true,
"groups": 9,
"durationMs": 38865,
"evidence": "docs/milestone-audit-delivery-ui-validation.json",
"log": ".qa/milestone-audit-verify-delivery-ui.log"
},
{
"script": "verify-index-ui.mjs",
"success": true,
"groups": 8,
"durationMs": 5702,
"evidence": "docs/milestone-audit-index-ui-validation.json",
"log": ".qa/milestone-audit-verify-index-ui.log"
},
{
"script": "verify-change-sets-ui.mjs",
"success": true,
"groups": 11,
"durationMs": 10331,
"evidence": "docs/milestone-audit-change-sets-ui-validation.json",
"log": ".qa/milestone-audit-verify-change-sets-ui.log"
},
{
"script": "verify-context-ui.mjs",
"success": true,
"groups": 11,
"durationMs": 16116,
"evidence": "docs/milestone-audit-context-ui-validation.json",
"log": ".qa/milestone-audit-verify-context-ui.log"
},
{
"script": "verify-workspaces-ui.mjs",
"success": true,
"groups": 9,
"durationMs": 20712,
"evidence": "docs/milestone-audit-workspaces-ui-validation.json",
"log": ".qa/milestone-audit-verify-workspaces-ui.log"
},
{
"script": "verify-python-ui.mjs",
"success": true,
"groups": 7,
"durationMs": 17306,
"evidence": "docs/milestone-audit-python-ui-validation.json",
"log": ".qa/milestone-audit-verify-python-ui.log"
},
{
"script": "verify-react-delivery.mjs",
"success": true,
"groups": 6,
"durationMs": 20241,
"evidence": "docs/milestone-audit-react-delivery-validation.json",
"log": ".qa/milestone-audit-verify-react-delivery.log"
},
{
"script": "verify-research-ui.mjs",
"success": true,
"groups": 11,
"durationMs": 20732,
"evidence": "docs/milestone-audit-research-ui-validation.json",
"log": ".qa/milestone-audit-verify-research-ui.log"
},
{
"script": "verify-engines-ui.mjs",
"success": true,
"groups": 10,
"durationMs": 9639,
"evidence": "docs/milestone-audit-engines-ui-validation.json",
"log": ".qa/milestone-audit-verify-engines-ui.log"
},
{
"script": "verify-model-center-ui.mjs",
"success": true,
"groups": 5,
"durationMs": 4232,
"evidence": "docs/milestone-audit-model-center-ui-validation.json",
"log": ".qa/milestone-audit-verify-model-center-ui.log"
},
{
"script": "verify-project-memory-ui.mjs",
"success": true,
"groups": 4,
"durationMs": 5398,
"evidence": "docs/milestone-audit-project-memory-ui-validation.json",
"log": ".qa/milestone-audit-verify-project-memory-ui.log"
},
{
"script": "verify-task-checkpoints-ui.mjs",
"success": true,
"groups": 3,
"durationMs": 5983,
"evidence": "docs/milestone-audit-task-checkpoints-ui-validation.json",
"log": ".qa/milestone-audit-verify-task-checkpoints-ui.log"
},
{
"script": "verify-skills-ui.mjs",
"success": true,
"groups": 4,
"durationMs": 6365,
"evidence": "docs/milestone-audit-skills-ui-validation.json",
"log": ".qa/milestone-audit-verify-skills-ui.log"
},
{
"script": "verify-mcp-ui.mjs",
"success": true,
"groups": 6,
"durationMs": 9265,
"evidence": "docs/milestone-audit-mcp-ui-validation.json",
"log": ".qa/milestone-audit-verify-mcp-ui.log"
},
{
"script": "verify-mcp-recovery-ui.mjs",
"success": true,
"groups": 2,
"durationMs": 5636,
"evidence": "docs/milestone-audit-mcp-recovery-ui-validation.json",
"log": ".qa/milestone-audit-verify-mcp-recovery-ui.log"
}
],
"scope": "Every verifier owns an independent port and synthetic data directory. Synthetic provider fixtures verify protocols; configured-model acceptance is separate. No production service or real key read/copied."
}
@@ -0,0 +1,27 @@
{
"success": true,
"url": "http://127.0.0.1:62044",
"browser": "Microsoft Edge 156.0.4314.50",
"date": "2026-10-30T15:33:22.785Z",
"projectRoot": "D:\\Ai-project\\REA逆向\\Agent-Workbench\\.qa\\ui-e91e622c-b16b-41c3-a42f-e139034db17e\\ui-project",
"viewports": [
1440,
780,
390
],
"passed": [
"project selection and real file preview",
"real command consent, execution and stale snapshot rejection",
"diff review and verified file application",
"manual editor creates a reviewable proposal",
"file responses stay with the current selection",
"denial, session selection, reload, stale response rejection and export",
"project inventory responses cannot replace another project",
"780 and 390 px: project drawer, file, diff and trace controls, no page overflow",
"unified and detailed research reports render",
"model list, ephemeral key, compatible fixture tool loop and return to offline mode"
],
"pageErrors": [],
"requestFailures": [],
"modelValidation": "controlled local HTTP fixture; no real model"
}
@@ -0,0 +1,29 @@
{
"success": true,
"version": "1.12.1",
"testedAt": "2026-10-30T15:35:02.875Z",
"cases": [
"Desktop batch scope, per-file diffs, sealed single-write guard and pre-apply bytes",
"Three-file reviewed application preserves exact BOM/CRLF/LF and creates the new file",
"Explicit batch rollback restores exact originals and removes only the reviewed new file",
"A last-file owner conflict blocks the whole batch and shows the real recovery direction",
"Durable partial recovery reads actual hashes and preserves owner content during rollback",
"Changed observation fingerprints reject stale acknowledgement without releasing the project",
"Explicit owner-content acknowledgement ends recovery without claiming successful application",
"Switching projects clears the previous task notification while retaining its durable error and recovery history",
"390px step-limit review, same-task continuation entry, rollback confirmation and keyboard close fit the viewport",
"A delayed preparation response never opens a different project or changes its selected session",
"A failed batch response after its component unmounts cannot display an error in the newly selected project"
],
"pageErrors": [],
"requestFailures": [],
"screenshots": [
".qa/screenshots/milestone-audit-batch-desktop-review.png",
".qa/screenshots/milestone-audit-batch-desktop-conflict.png",
".qa/screenshots/milestone-audit-batch-desktop-recovery.png",
".qa/screenshots/milestone-audit-batch-mobile-review.png",
".qa/screenshots/milestone-audit-batch-mobile-rollback.png"
],
"projectRoot": "D:\\Ai-project\\REA逆向\\Agent-Workbench\\.qa\\batch-ui-d2725d6a-4d2b-4023-bac5-8e539be91e34",
"scope": "Real independent demo host and durable batch states; UI fixtures include actual owner conflict/rollback results. No real model credentials or user project were used."
}
@@ -0,0 +1,175 @@
{
"success": true,
"version": "1.12.1",
"testedAt": "2026-10-30T15:33:03.495Z",
"directory": "D:\\Ai-project\\REA逆向\\Agent-Workbench\\.qa\\browser-ui-f085de99-e76f-42b4-8a57-e58b64852e15",
"passed": [
"actual step editor seals the chosen build, viewport and three CSS interactions without opening the page",
"review shows exact build identity and planned assertions; cancelling executes no browser work",
"real inner Edge clicks the intended element and the two counted assertions display with original Task/Run/build identity",
"original large JSON is paged in the UI and downloads the complete hash-identical result",
"the original screenshot renders at its actual resolution and downloads identical PNG bytes",
"a real failed assertion retains actual count one and requires a newly reviewed plan instead of replaying",
"390px forms and result records remain usable and the inner browser captures its real 390px layout",
"stopping an actual in-flight assertion closes its owned browser and preserves prior successful and failed jobs",
"a delayed original-record failure stays with its source and does not appear in a newly selected task",
"actual host restart preserves the four original browser jobs and PNGs without model requests or execution replay"
],
"pageErrors": [],
"requestFailures": [],
"screenshots": [
"D:\\Ai-project\\REA逆向\\Agent-Workbench\\.qa\\browser-ui-f085de99-e76f-42b4-8a57-e58b64852e15\\browser-desktop-results.png",
"D:\\Ai-project\\REA逆向\\Agent-Workbench\\.qa\\browser-ui-f085de99-e76f-42b4-8a57-e58b64852e15\\browser-desktop-original-image.png",
"D:\\Ai-project\\REA逆向\\Agent-Workbench\\.qa\\browser-ui-f085de99-e76f-42b4-8a57-e58b64852e15\\browser-mobile-results.png",
"D:\\Ai-project\\REA逆向\\Agent-Workbench\\.qa\\browser-ui-f085de99-e76f-42b4-8a57-e58b64852e15\\browser-mobile-original-image.png"
],
"taskId": "18f51708-f98f-4028-b0f0-d528517d004b",
"sessionId": "88e522fb-75be-479e-b1a8-a7540153552d",
"jobs": [
{
"id": "deaced0d-8a49-4cec-9c72-efe447265f51",
"status": "passed",
"planId": "12ed8d3c-831c-48c6-b855-45ca0106c9db",
"assertions": {
"total": 2,
"passed": 2,
"failed": 0
},
"resultSha256": "02097a77a6ad9a55bf4285e08577c2fd402a111ce6c889647e93240572c55227",
"cleanup": {
"pid": 33724,
"exited": true,
"exitCode": 0,
"signal": null,
"finishedAt": "2026-10-30T15:32:48.847Z",
"previewClosed": true,
"snapshotDisposed": true
},
"screenshots": [
{
"artifactId": "e9ac2479-943a-457b-978e-9bfb7e738217",
"sha256": "74f2387b837f5f74bd7e8eef7631865a3b5bb24601c4bbaa93b2478a9e94c6ef",
"bytes": 101061,
"width": 1280,
"height": 900,
"label": "initial"
},
{
"artifactId": "45e34136-62af-42ab-98e4-48d8639f3dde",
"sha256": "b1de525bcf042bc1724e8a9c6670042ab3035bed394ff76cf2a934a443ae98f7",
"bytes": 85261,
"width": 1280,
"height": 900,
"label": "final"
}
]
},
{
"id": "0dffbb92-c53b-4e13-92d8-eb4891887d94",
"status": "failed",
"planId": "70bf237a-a3e6-49d3-a14d-b85428731d0c",
"assertions": {
"total": 2,
"passed": 1,
"failed": 1
},
"resultSha256": "78f972e9998e45671c72873570da7e4b797a4b423b7cc821d8d6dd10a03feb26",
"cleanup": {
"pid": 31476,
"exited": true,
"exitCode": 0,
"signal": null,
"finishedAt": "2026-10-30T15:32:55.947Z",
"previewClosed": true,
"snapshotDisposed": true
},
"screenshots": [
{
"artifactId": "f24f2aad-e1d2-4c16-9afe-8da0bb26fc75",
"sha256": "74f2387b837f5f74bd7e8eef7631865a3b5bb24601c4bbaa93b2478a9e94c6ef",
"bytes": 101061,
"width": 1280,
"height": 900,
"label": "initial"
},
{
"artifactId": "5c76f367-f893-44ba-a911-7dd3d6228326",
"sha256": "b1de525bcf042bc1724e8a9c6670042ab3035bed394ff76cf2a934a443ae98f7",
"bytes": 85261,
"width": 1280,
"height": 900,
"label": "failure"
}
]
},
{
"id": "6c5a11f7-f777-45c6-9508-3af32e3c04ba",
"status": "passed",
"planId": "248491fa-e1f7-4813-b672-fe8c7a2fb236",
"assertions": {
"total": 2,
"passed": 2,
"failed": 0
},
"resultSha256": "ef94eb9b8d4a4b38508e076cfbe3e10d76be2ef78052b16c760699a1dee9553e",
"cleanup": {
"pid": 44264,
"exited": true,
"exitCode": 0,
"signal": null,
"finishedAt": "2026-10-30T15:32:58.384Z",
"previewClosed": true,
"snapshotDisposed": true
},
"screenshots": [
{
"artifactId": "b8f1e9e6-ea64-486d-b95a-5f7a0a244fde",
"sha256": "93969f1954eec1f7fa4736d7a3035631c6066b6727132ff1179180eadb745602",
"bytes": 11011,
"width": 390,
"height": 844,
"label": "initial"
},
{
"artifactId": "3e5da9ee-d515-4f23-88a7-77234bf884eb",
"sha256": "a6908ecfcfd239721ec6bbb916a5054d2f61183b6dea3f014db89216d371d0e3",
"bytes": 18832,
"width": 390,
"height": 844,
"label": "final"
}
]
},
{
"id": "16bfde86-5828-4bef-9f8b-3dc4979e6012",
"status": "interrupted",
"planId": "401d5a66-74e9-40dd-b46f-d168e56e6da4",
"assertions": {
"total": 3,
"passed": 2,
"failed": 1
},
"resultSha256": "b6b6fb35eed88b5fcf6b0dcdf219877d0b18eb315700d39a481de634dbd1e206",
"cleanup": {
"pid": 43720,
"exited": true,
"exitCode": 0,
"signal": null,
"finishedAt": "2026-10-30T15:33:01.561Z",
"previewClosed": true,
"snapshotDisposed": true
},
"screenshots": [
{
"artifactId": "93392adf-5be2-4a45-99dd-f6966dd09d57",
"sha256": "93969f1954eec1f7fa4736d7a3035631c6066b6727132ff1179180eadb745602",
"bytes": 11011,
"width": 390,
"height": 844,
"label": "initial"
}
]
}
],
"scope": "Actual independent Workbench UI plus real Edge static-build executions. The configured production Provider, other projects and keys are not accessed."
}
@@ -0,0 +1,29 @@
{
"success": true,
"version": "1.12.1",
"testedAt": "2026-10-30T15:35:02.875Z",
"cases": [
"Desktop batch scope, per-file diffs, sealed single-write guard and pre-apply bytes",
"Three-file reviewed application preserves exact BOM/CRLF/LF and creates the new file",
"Explicit batch rollback restores exact originals and removes only the reviewed new file",
"A last-file owner conflict blocks the whole batch and shows the real recovery direction",
"Durable partial recovery reads actual hashes and preserves owner content during rollback",
"Changed observation fingerprints reject stale acknowledgement without releasing the project",
"Explicit owner-content acknowledgement ends recovery without claiming successful application",
"Switching projects clears the previous task notification while retaining its durable error and recovery history",
"390px step-limit review, same-task continuation entry, rollback confirmation and keyboard close fit the viewport",
"A delayed preparation response never opens a different project or changes its selected session",
"A failed batch response after its component unmounts cannot display an error in the newly selected project"
],
"pageErrors": [],
"requestFailures": [],
"screenshots": [
".qa/screenshots/milestone-audit-batch-desktop-review.png",
".qa/screenshots/milestone-audit-batch-desktop-conflict.png",
".qa/screenshots/milestone-audit-batch-desktop-recovery.png",
".qa/screenshots/milestone-audit-batch-mobile-review.png",
".qa/screenshots/milestone-audit-batch-mobile-rollback.png"
],
"projectRoot": "D:\\Ai-project\\REA逆向\\Agent-Workbench\\.qa\\batch-ui-d2725d6a-4d2b-4023-bac5-8e539be91e34",
"scope": "Real independent demo host and durable batch states; UI fixtures include actual owner conflict/rollback results. No real model credentials or user project were used."
}
File diff suppressed because it is too large Load Diff
@@ -0,0 +1,34 @@
{
"success": true,
"version": "1.12.1",
"testedAt": "2026-10-30T15:35:18.503Z",
"passed": [
"desktop request estimate, actual usage, complete goal and history identity are visible",
"paged complete history opens actual tool result artifacts without executing tools",
"historical and resumed Runs retain their own request records",
"390px history, original JSON and sent-body dialogs fit and retain close controls",
"late old history failure stays in its original scope after switching projects",
"late old Run failure stays in its original scope after switching projects",
"late old request-body failure stays in its original scope after switching projects",
"context-limit pause explains the intact requirement and explicit continuation",
"390px repeated-read pause explains the cause and explicit continuation retains the same Task",
"desktop working evidence shows actual code, original result provenance and the current Task",
"390px working evidence and sent request stay readable and closing the artifact retains task navigation"
],
"pageErrors": [],
"requestFailures": [],
"screenshots": [
".qa/screenshots/milestone-audit-desktop-plan.png",
".qa/screenshots/milestone-audit-desktop-history.png",
".qa/screenshots/milestone-audit-mobile-history.png",
".qa/screenshots/milestone-audit-mobile-original.png",
".qa/screenshots/milestone-audit-mobile-request-body.png",
".qa/screenshots/milestone-audit-mobile-read-pause.png",
".qa/screenshots/milestone-audit-desktop-working-evidence.png",
".qa/screenshots/milestone-audit-mobile-working-evidence.png"
],
"root": "D:\\Ai-project\\REA逆向\\Agent-Workbench\\.qa\\context-ui-f82e3e7d-f745-450c-aca6-f9e36f3ab43d",
"data": "D:\\Ai-project\\REA逆向\\Agent-Workbench\\.qa\\context-ui-f82e3e7d-f745-450c-aca6-f9e36f3ab43d\\data",
"mode": "saved synthetic context evidence; no real model or project commands executed by this UI verifier",
"browser": "156.0.4314.50"
}
@@ -0,0 +1,31 @@
{
"success": true,
"version": "1.12.1",
"testedAt": "2026-10-30T15:34:46.656Z",
"browser": "156.0.4314.50",
"passed": [
"actual failing tests and four source proposals retain the original project and task",
"reviewed changes continue the same Task through 14 real tests and model-prepared Git review",
"review shows actual paths, diffs, tests, author and policy; cancelling leaves Git unchanged",
"an index change after confirmation preview refuses the stale job and preserves both staged owners",
"390px confirmation keeps file hashes, diffs and actions inside the modal",
"double confirmation creates one four-file commit, retains owner bytes and publishes report with completion",
"report viewing and Markdown/JSON downloads preserve verified artifact bytes and truthful acceptance limits",
"mobile report preview closes cleanly and keeps the task input fixed",
"unknown Git results block new submission and the explicit reconcile action restores the existing commit"
],
"pageErrors": [],
"requestFailures": [],
"projectRoot": "D:\\Ai-project\\REA逆向\\Agent-Workbench\\.qa\\delivery-ui-74712e04-69ea-4a10-a449-bb1b393fb957\\四模块交付工程",
"projectId": "d18c3b65-68ff-4167-96e1-7790bc5bf360",
"sessionId": "de562ac7-8f05-4dd3-a0d4-5590c71b805b",
"taskId": "7afcdf67-c120-4c45-b045-4bddca74bbe7",
"commit": "d67e17670937ab0a577527a6790eb31e23ac8f95",
"recoveryCommit": "67d0de7aba5c67ef627090cf8e5e1b1f638d5ae2",
"testCounts": {
"baselineFailed": 13,
"passed": 14
},
"model": "controlled local HTTP fixture",
"confirmations": 2
}
@@ -0,0 +1,26 @@
{
"testedAt": "2026-10-30T14:11:39.966Z",
"command": "npm audit --json --ignore-scripts",
"exitCode": 0,
"metadata": {
"vulnerabilities": {
"info": 0,
"low": 0,
"moderate": 0,
"high": 0,
"critical": 0,
"total": 0
},
"dependencies": {
"prod": 13,
"dev": 141,
"optional": 79,
"peer": 0,
"peerOptional": 0,
"total": 153
}
},
"error": null,
"outputSha256": "fb8761bffb39edf97e113a34db5f3d673ed9354a74251aa3a31255462d8ad025",
"findings": []
}
@@ -0,0 +1,20 @@
{
"success": true,
"version": "1.12.1",
"url": "http://127.0.0.1:57675",
"date": "2026-10-30T15:33:42.571Z",
"passed": [
"model response validation checks the tentative connection before saving",
"a model failure after successful tools shows source, attempts and local runtime diagnosis",
"explicit retry preserves the completed tool result without replaying the task",
"an error question receives the actual session diagnosis and formatted answer",
"review state exposes the review action and prevents application during a running request",
"approved Unicode commands execute in the declared shell and retain Chinese output",
"authentication errors do not retry and error/diagnostic controls fit a 390 px viewport",
"stop cancels pending retry and clears running model state",
"a transient service failure recovers within the bounded retry policy"
],
"pageErrors": [],
"requestFailures": [],
"modelValidation": "controlled local HTTP fixture; actual configured model is verified separately"
}
@@ -0,0 +1,232 @@
{
"success": true,
"version": "1.12.1",
"testedAt": "2026-10-30T15:36:47.681Z",
"root": "D:\\Ai-project\\REA逆向\\Agent-Workbench\\.qa\\engines-ui-cbc842de-b91c-49e2-b603-144d8dfc57ae",
"browser": "156.0.4314.50",
"passed": [
"independent desktop engine entry detects the declared installation and clearly separates connection from project execution",
"missing installations and edited directories cannot launch a check until detection succeeds",
"UI launch seals the selected installation into the current Task/Run and renders actual startup and process-close facts",
"original record paging, full download SHA256 and nested-dialog keyboard close preserve the exact evidence",
"an explicit new check creates a different process and record while retaining the original check unchanged",
"declared pending-initialize fixture stops its owned process and explicit retry preserves the stopped result",
"the same product UI verifies the actual installed DeepSeek SDK without a prompt or configured-model call",
"390px installation paths, source chain, original record, paging and keyboard close remain readable without page overflow",
"late detection failure from a closed panel cannot appear in a different task or reveal its prior engine records",
"actual host restart preserves all native and declared-fixture records with no protocol replay or project modification"
],
"pageErrors": [],
"requestFailures": [],
"expectedRequestFailures": [],
"screenshots": [
"D:\\Ai-project\\REA逆向\\Agent-Workbench\\.qa\\engines-ui-cbc842de-b91c-49e2-b603-144d8dfc57ae\\screenshots\\engines-native-desktop.png",
"D:\\Ai-project\\REA逆向\\Agent-Workbench\\.qa\\engines-ui-cbc842de-b91c-49e2-b603-144d8dfc57ae\\screenshots\\engines-record-mobile.png",
"D:\\Ai-project\\REA逆向\\Agent-Workbench\\.qa\\engines-ui-cbc842de-b91c-49e2-b603-144d8dfc57ae\\screenshots\\engines-native-mobile.png"
],
"starts": [
{
"root": "D:\\Ai-project\\REA逆向\\Agent-Workbench\\.qa\\engines-ui-cbc842de-b91c-49e2-b603-144d8dfc57ae\\本地 SDK #",
"identitySha256": "e9b0fb8df3c1b703b020d144f1c01c48b6eb6c896104f12a39b5a5fadb3b615f"
},
{
"root": "D:\\Ai-project\\REA逆向\\Agent-Workbench\\.qa\\engines-ui-cbc842de-b91c-49e2-b603-144d8dfc57ae\\本地 SDK #",
"retry": true
},
{
"root": "D:\\Ai-project\\REA逆向\\Agent-Workbench\\.qa\\engines-ui-cbc842de-b91c-49e2-b603-144d8dfc57ae\\本地 SDK #",
"identitySha256": "e9b0fb8df3c1b703b020d144f1c01c48b6eb6c896104f12a39b5a5fadb3b615f"
},
{
"root": "D:\\Ai-project\\REA逆向\\Agent-Workbench\\.qa\\engines-ui-cbc842de-b91c-49e2-b603-144d8dfc57ae\\本地 SDK #",
"retry": true
},
{
"root": "D:\\Ai-project\\DeepSeek-Harness-Desktop\\deepseek-harness",
"identitySha256": "a60f7733d6eaf91bcd31ce5d1da38d6d267a16ad29bc03bba548f314fddec22a"
}
],
"nativeJob": {
"schemaVersion": 1,
"id": "007ca020-bbfb-482d-8357-1f39f21d4f3e",
"sessionId": "6de0e0b8-d03b-43a5-901b-5cee65114430",
"projectId": "a88c3e16-e6ea-4d75-9187-bc3140c37e42",
"taskId": "bfed1813-8b7a-44fc-881a-23bdccf7ac2e",
"runId": "2da1ea69-3436-45b0-8b87-05afbf8e69a2",
"engineId": "deepseek-harness",
"scope": "deepseek-sdk-connection-probe",
"source": "user",
"installation": {
"engineId": "deepseek-harness",
"name": "DeepSeek Harness",
"profile": "sdk-minimal",
"expectedVersion": "0.2.1-rc.2",
"root": "D:\\Ai-project\\DeepSeek-Harness-Desktop\\deepseek-harness",
"capabilities": {
"connectionProbe": true,
"projectExecution": false,
"prompt": false,
"singleSessionCancel": false,
"protocolVersionNegotiation": false,
"cancellationUnit": "owned-process",
"tokenStreaming": false
},
"limitations": [
"当前接入核验启动、会话目录、协议错误和关闭;工程执行接入开发中。",
"SDK没有单会话cancel或close;停止以所属独立进程为单位。",
"SDK serverInfo.version与发行包版本分别核对,没有有效协议版本协商。",
"SDK事件覆盖运行时会话树,idle事件不能直接判项目交付完成。"
],
"status": "available",
"version": "0.2.1-rc.2",
"entry": "D:\\Ai-project\\DeepSeek-Harness-Desktop\\deepseek-harness\\apps\\cli\\lib\\bin.js",
"inputs": [
{
"path": "D:\\Ai-project\\DeepSeek-Harness-Desktop\\deepseek-harness\\apps\\cli\\package.json",
"bytes": 8138,
"sha256": "d42efe9f3ff424f602f11598af8805ad25039953ebbbd3c163d67875f7e8658b",
"relative": "apps/cli/package.json"
},
{
"path": "D:\\Ai-project\\DeepSeek-Harness-Desktop\\deepseek-harness\\apps\\cli\\lib\\bin.js",
"bytes": 11772,
"sha256": "80112efe368efb1a39720c7ada0af00442e5131fb2744edaaff8a3e8bb499e6f",
"relative": "apps/cli/lib/bin.js"
},
{
"path": "D:\\Ai-project\\DeepSeek-Harness-Desktop\\deepseek-harness\\packages\\boot\\cmdline\\lib\\index.js",
"bytes": 6973,
"sha256": "7253190611566485f8b6f836fae6e0f2a7abbc59bb807c42e874844016e6d3a0",
"relative": "packages/boot/cmdline/lib/index.js"
},
{
"path": "D:\\Ai-project\\DeepSeek-Harness-Desktop\\deepseek-harness\\packages\\bundle\\sdk-app\\lib\\index.js",
"bytes": 1921,
"sha256": "cc7e6c38b0b38c0e92a0dffb45210c1dac6ef518547436e805e84b772f91a475",
"relative": "packages/bundle/sdk-app/lib/index.js"
},
{
"path": "D:\\Ai-project\\DeepSeek-Harness-Desktop\\deepseek-harness\\packages\\bundle\\sdk-minimal\\package.json",
"bytes": 2920,
"sha256": "0e177cbca85148207343db5afd7f12829723cf3f2afedfa8e85a5c5dcd58f604",
"relative": "packages/bundle/sdk-minimal/package.json"
},
{
"path": "D:\\Ai-project\\DeepSeek-Harness-Desktop\\deepseek-harness\\packages\\bundle\\sdk-minimal\\cordis.patch.yml",
"bytes": 5636,
"sha256": "464ea07c03d3639c2c387fa3a684f5d4001e7ca39f8257b361c604f19487af66",
"relative": "packages/bundle/sdk-minimal/cordis.patch.yml"
},
{
"path": "D:\\Ai-project\\DeepSeek-Harness-Desktop\\deepseek-harness\\packages\\sdk\\server\\package.json",
"bytes": 2155,
"sha256": "2690f99763c3e1e7bd57f680f26e53010c2a3d7c4dffe06ff64a07ae3928cb17",
"relative": "packages/sdk/server/package.json"
},
{
"path": "D:\\Ai-project\\DeepSeek-Harness-Desktop\\deepseek-harness\\packages\\sdk\\server\\lib\\index.js",
"bytes": 16454,
"sha256": "55412923f9f7c279f7e0b2a1e2853eaf238f390b3d84046cb9c94dc5f22a1e4a",
"relative": "packages/sdk/server/lib/index.js"
},
{
"path": "D:\\Ai-project\\DeepSeek-Harness-Desktop\\deepseek-harness\\packages\\sdk\\protocol\\package.json",
"bytes": 1237,
"sha256": "b030b0431d69723b5424dc2519cd4b381ec91acd22b73e0b30d6b8a19f1f2f94",
"relative": "packages/sdk/protocol/package.json"
},
{
"path": "D:\\Ai-project\\DeepSeek-Harness-Desktop\\deepseek-harness\\packages\\sdk\\protocol\\lib\\index.js",
"bytes": 7336,
"sha256": "371c38ca3514797d86df5f4bb78a56c40a2637294b83aababe8145bec6a20253",
"relative": "packages/sdk/protocol/lib/index.js"
},
{
"path": "D:\\Ai-project\\DeepSeek-Harness-Desktop\\deepseek-harness\\packages\\sdk\\client\\package.json",
"bytes": 1325,
"sha256": "ff33f2ecec2d48bb8a24397cc664fc2dd50ba49d5288eb11fea1cb12931e5d5f",
"relative": "packages/sdk/client/package.json"
},
{
"path": "D:\\Ai-project\\DeepSeek-Harness-Desktop\\deepseek-harness\\pnpm-lock.yaml",
"bytes": 1090222,
"sha256": "4637cca3b89c4c8d5a147321710e597aa62ec2444a3772ab542be5690893dfe4",
"relative": "pnpm-lock.yaml"
},
{
"path": "D:\\Ai-project\\DeepSeek-Harness-Desktop\\deepseek-harness\\vendor\\cordis\\lib\\index.js",
"bytes": 60361,
"sha256": "6a9394c0877ff45218818c6e815edd038f8057e1a1deb390a8d43ec81c57691e",
"relative": "vendor/cordis/lib/index.js"
}
],
"node": {
"path": "C:\\Program Files\\nodejs\\node.exe",
"bytes": 93915464,
"sha256": "d702daa815eabbff763c7d3808e235cf031cf2ca70211b56847bc72f51e240e5",
"version": "24.22.1"
},
"identitySha256": "a60f7733d6eaf91bcd31ce5d1da38d6d267a16ad29bc03bba548f314fddec22a",
"serverInfo": {
"name": "deepseek-harness-sdk-runtime",
"version": "0.0.1"
},
"message": "已找到匹配版本,可验证原程序的连接与停止。"
},
"sdkSessionId": "9db6910d-75d0-41cd-a446-861406f2417b",
"directory": "C:\\Users\\wangy\\AppData\\Local\\Temp\\AgentWorkbench-Engines\\007ca020-bbfb-482d-8357-1f39f21d4f3e",
"status": "completed",
"startedAt": "2026-10-30T15:36:44.645Z",
"finishedAt": "2026-10-30T15:36:45.705Z",
"requiresVerification": false,
"verification": {
"verified": true,
"installationSha256": "a60f7733d6eaf91bcd31ce5d1da38d6d267a16ad29bc03bba548f314fddec22a",
"readySignal": "appReady",
"sdkSessionId": "9db6910d-75d0-41cd-a446-861406f2417b",
"connectionId": "8817a1b8-9253-4364-96ec-2fdc5157ff36",
"rootExited": true,
"modelRequests": 0
},
"summary": "连接通过,启动、协议返回、会话目录与进程结束均已核对。",
"contractArtifactId": "6c4abd20-07e6-45ac-9dd1-724bd14e60b6",
"contractSha256": "e03e183842b77c48211adb14410f49b484eba9f2b2b846f9ab762f6ce4194b29",
"eventId": "4ca0b22b-1e6a-416a-bdbc-63ab0ea4dac5",
"pid": 41192,
"connectionId": "8817a1b8-9253-4364-96ec-2fdc5157ff36",
"readiness": {
"schemaVersion": 1,
"nonce": "d31c8e64-f0e9-424a-be17-faa74f27bda4",
"pid": 41192,
"signal": "appReady",
"readyAt": "2026-10-30T15:36:45.505Z",
"artifact": {
"path": "C:\\Users\\wangy\\AppData\\Local\\Temp\\AgentWorkbench-Engines\\007ca020-bbfb-482d-8357-1f39f21d4f3e\\ready.json",
"bytes": 135,
"sha256": "c34502c77b73eb915f0eaef17b5050fd53a4d9bcfdc26581ff6eaa2413747a53"
}
},
"serverInfo": {
"name": "deepseek-harness-sdk-runtime",
"version": "0.0.1"
},
"reportArtifactId": "d25e6269-9c27-4135-9bfe-d9e20a084c25",
"reportSha256": "cf9718d81e153f4d337627ff63e0fbc6c45b556c8c866bc117f3aafc0f2af480",
"reportSummary": {
"status": "verified",
"modelRequests": 0,
"providerRequests": 0,
"observedResponses": 6,
"rootExited": true
}
},
"actualSdk": {
"version": "0.2.1-rc.2",
"identitySha256": "a60f7733d6eaf91bcd31ce5d1da38d6d267a16ad29bc03bba548f314fddec22a",
"nativeMethods": 6,
"modelRequests": 0,
"rootExited": true
},
"stopScope": "declared private SDK pending-initialize fixture",
"scope": "Independent service and data. Original 4318 Provider, tasks and credentials were not accessed."
}
@@ -0,0 +1,28 @@
{
"at": "2026-10-30T14:30:44.640Z",
"directory": "D:\\Ai-project\\REA逆向\\Agent-Workbench\\.qa\\milestone-environment-tR5epo",
"shell": {
"status": "completed",
"inherited": {
"OPENAI_API_KEY": true,
"MODEL_ROUTER_TOKEN": true,
"CUSTOM_SERVICE_SECRET": true
}
},
"npm": {
"status": "passed",
"inherited": {
"OPENAI_API_KEY": true,
"MODEL_ROUTER_TOKEN": true,
"CUSTOM_SERVICE_SECRET": true
}
},
"python": {
"inherited": {
"OPENAI_API_KEY": true,
"MODEL_ROUTER_TOKEN": true,
"CUSTOM_SERVICE_SECRET": true
}
},
"realCredentials": "none; worker receives only system paths and owned markers"
}
@@ -0,0 +1,19 @@
{
"success": true,
"version": "1.12.1",
"testedAt": "2026-10-30T15:34:07.743Z",
"browser": "156.0.4314.50",
"passed": [
"model-selected checks wait for approval and display actual counted results",
"verified artifact preview shows the real check log and hash",
"changed source invalidates old checks and manual confirmation runs the new plan",
"390px task drawer, log preview, close controls and fixed composer remain operable",
"reconnection synchronizes missed session changes in other projects",
"budget controls pause before tools and preserve the goal across recorded runs",
"applied proposals continue the original task into actual engineering verification"
],
"pageErrors": [],
"requestFailures": [],
"projectRoot": "D:\\Ai-project\\REA逆向\\Agent-Workbench\\.qa\\execution-ui-80ed42a5-0f92-43fa-8eda-2a60bbf07314\\project",
"model": "controlled local HTTP fixture"
}
@@ -0,0 +1,264 @@
{
"codeCommit": "f4bd0ad",
"releaseId": "20261030150404738",
"uiGroups": 150,
"sourceSuccess": false,
"sourceTests": {
"tests": 369,
"pass": 368,
"fail": 1,
"cancelled": 0,
"skipped": 0,
"todo": 0
},
"originals": [
{
"path": "docs/milestone-audit-all-ui-validation.json",
"saved": "D:\\Ai-project\\REA逆向\\Agent-Workbench\\.qa\\milestone-first-candidate-3bd2cc18-321b-4a9e-800e-9ad645be72c2\\docs\\milestone-audit-all-ui-validation.json",
"sha256": "79f87769d56cfdef78c62078726c6b14005d4c330e8f19043892a54f57414168",
"bytes": 5409
},
{
"path": "docs/milestone-audit-task-ui-validation.json",
"saved": "D:\\Ai-project\\REA逆向\\Agent-Workbench\\.qa\\milestone-first-candidate-3bd2cc18-321b-4a9e-800e-9ad645be72c2\\docs\\milestone-audit-task-ui-validation.json",
"sha256": "1aaa5de3d630901056df0a81f67b3b73c87c7266ac982eb32cdc1aa470c38d85",
"bytes": 2303
},
{
"path": ".qa/milestone-audit-verify-task-plans-ui.log",
"saved": "D:\\Ai-project\\REA逆向\\Agent-Workbench\\.qa\\milestone-first-candidate-3bd2cc18-321b-4a9e-800e-9ad645be72c2\\.qa\\milestone-audit-verify-task-plans-ui.log",
"sha256": "40c7e0623140f96de42078471826acdaab74345c3964b2cb144d040d33bb6432",
"bytes": 993
},
{
"path": "docs/milestone-audit-browser-ui-validation.json",
"saved": "D:\\Ai-project\\REA逆向\\Agent-Workbench\\.qa\\milestone-first-candidate-3bd2cc18-321b-4a9e-800e-9ad645be72c2\\docs\\milestone-audit-browser-ui-validation.json",
"sha256": "897df7efc1d005712f5d999c00aeef9cb015a0c815116503b6b843d4ea6726d0",
"bytes": 6537
},
{
"path": ".qa/milestone-audit-verify-browser-ui.log",
"saved": "D:\\Ai-project\\REA逆向\\Agent-Workbench\\.qa\\milestone-first-candidate-3bd2cc18-321b-4a9e-800e-9ad645be72c2\\.qa\\milestone-audit-verify-browser-ui.log",
"sha256": "9e845542bfcfb421c30a872a46f5479ceb8fa83c3f94063716545a0e0f4e481f",
"bytes": 1235
},
{
"path": "docs/milestone-audit-basic-ui-validation.json",
"saved": "D:\\Ai-project\\REA逆向\\Agent-Workbench\\.qa\\milestone-first-candidate-3bd2cc18-321b-4a9e-800e-9ad645be72c2\\docs\\milestone-audit-basic-ui-validation.json",
"sha256": "ab10ccbb0d771728f851ec32a4d0ddec408d158330026e85a5a56ec4dbdafbcb",
"bytes": 1083
},
{
"path": ".qa/milestone-audit-verify-ui.log",
"saved": "D:\\Ai-project\\REA逆向\\Agent-Workbench\\.qa\\milestone-first-candidate-3bd2cc18-321b-4a9e-800e-9ad645be72c2\\.qa\\milestone-audit-verify-ui.log",
"sha256": "c5c4973a94fd979685551a97ac8ba0e6db3290331d60425f5ba1eebec06a4f8c",
"bytes": 1706
},
{
"path": "docs/milestone-audit-diagnostics-ui-validation.json",
"saved": "D:\\Ai-project\\REA逆向\\Agent-Workbench\\.qa\\milestone-first-candidate-3bd2cc18-321b-4a9e-800e-9ad645be72c2\\docs\\milestone-audit-diagnostics-ui-validation.json",
"sha256": "8ad68eaad9b19e8f325a7587d35972142f22273081e87dff2119e679eae482fa",
"bytes": 1047
},
{
"path": ".qa/milestone-audit-verify-diagnostics-ui.log",
"saved": "D:\\Ai-project\\REA逆向\\Agent-Workbench\\.qa\\milestone-first-candidate-3bd2cc18-321b-4a9e-800e-9ad645be72c2\\.qa\\milestone-audit-verify-diagnostics-ui.log",
"sha256": "f8c859409fac2a036041c2eebabe39cb86c98037fdf1ea16ed28b89d3c07178f",
"bytes": 1795
},
{
"path": "docs/milestone-audit-execution-ui-validation.json",
"saved": "D:\\Ai-project\\REA逆向\\Agent-Workbench\\.qa\\milestone-first-candidate-3bd2cc18-321b-4a9e-800e-9ad645be72c2\\docs\\milestone-audit-execution-ui-validation.json",
"sha256": "5dca13e15274d8017b3a71b52b82ad228a187554dae16c2c32ea1567a6058168",
"bytes": 924
},
{
"path": ".qa/milestone-audit-verify-execution-ui.log",
"saved": "D:\\Ai-project\\REA逆向\\Agent-Workbench\\.qa\\milestone-first-candidate-3bd2cc18-321b-4a9e-800e-9ad645be72c2\\.qa\\milestone-audit-verify-execution-ui.log",
"sha256": "f24594006927caeb0425b9ca6bfec1869ebda1c8742306d126a692752e86771c",
"bytes": 1484
},
{
"path": "docs/milestone-audit-delivery-ui-validation.json",
"saved": "D:\\Ai-project\\REA逆向\\Agent-Workbench\\.qa\\milestone-first-candidate-3bd2cc18-321b-4a9e-800e-9ad645be72c2\\docs\\milestone-audit-delivery-ui-validation.json",
"sha256": "ea2b5e458da4015fcedc471e704e8eb1e01251f0c49d424681636089af61da1e",
"bytes": 1622
},
{
"path": ".qa/milestone-audit-verify-delivery-ui.log",
"saved": "D:\\Ai-project\\REA逆向\\Agent-Workbench\\.qa\\milestone-first-candidate-3bd2cc18-321b-4a9e-800e-9ad645be72c2\\.qa\\milestone-audit-verify-delivery-ui.log",
"sha256": "94138e47ce181eff372e06a5d072c6b2c91c1270b9880eb8b04425d0377398e4",
"bytes": 2493
},
{
"path": "docs/milestone-audit-index-ui-validation.json",
"saved": "D:\\Ai-project\\REA逆向\\Agent-Workbench\\.qa\\milestone-first-candidate-3bd2cc18-321b-4a9e-800e-9ad645be72c2\\docs\\milestone-audit-index-ui-validation.json",
"sha256": "b7fa50d9fad9152d7fefd50ba6ada520daf0da17b7415b8837aa476ec3ccd1cc",
"bytes": 1108
},
{
"path": ".qa/milestone-audit-verify-index-ui.log",
"saved": "D:\\Ai-project\\REA逆向\\Agent-Workbench\\.qa\\milestone-first-candidate-3bd2cc18-321b-4a9e-800e-9ad645be72c2\\.qa\\milestone-audit-verify-index-ui.log",
"sha256": "b7fa50d9fad9152d7fefd50ba6ada520daf0da17b7415b8837aa476ec3ccd1cc",
"bytes": 1108
},
{
"path": "docs/milestone-audit-change-sets-ui-validation.json",
"saved": "D:\\Ai-project\\REA逆向\\Agent-Workbench\\.qa\\milestone-first-candidate-3bd2cc18-321b-4a9e-800e-9ad645be72c2\\docs\\milestone-audit-change-sets-ui-validation.json",
"sha256": "60e452263e54872b03b87b87c20377b2bbb48222c6d61ed117182aaac84fbc21",
"bytes": 1913
},
{
"path": ".qa/milestone-audit-verify-change-sets-ui.log",
"saved": "D:\\Ai-project\\REA逆向\\Agent-Workbench\\.qa\\milestone-first-candidate-3bd2cc18-321b-4a9e-800e-9ad645be72c2\\.qa\\milestone-audit-verify-change-sets-ui.log",
"sha256": "60e452263e54872b03b87b87c20377b2bbb48222c6d61ed117182aaac84fbc21",
"bytes": 1913
},
{
"path": "docs/milestone-audit-context-ui-validation.json",
"saved": "D:\\Ai-project\\REA逆向\\Agent-Workbench\\.qa\\milestone-first-candidate-3bd2cc18-321b-4a9e-800e-9ad645be72c2\\docs\\milestone-audit-context-ui-validation.json",
"sha256": "37aa07bb4e8d03d1291efcf7ecea483fad7567a8c390ea57d0be45c7944af857",
"bytes": 2019
},
{
"path": ".qa/milestone-audit-verify-context-ui.log",
"saved": "D:\\Ai-project\\REA逆向\\Agent-Workbench\\.qa\\milestone-first-candidate-3bd2cc18-321b-4a9e-800e-9ad645be72c2\\.qa\\milestone-audit-verify-context-ui.log",
"sha256": "46231edc520113f36edcfd63679ca0e3182b975b180d737008bf8ac8e4d7bf09",
"bytes": 2986
},
{
"path": "docs/milestone-audit-workspaces-ui-validation.json",
"saved": "D:\\Ai-project\\REA逆向\\Agent-Workbench\\.qa\\milestone-first-candidate-3bd2cc18-321b-4a9e-800e-9ad645be72c2\\docs\\milestone-audit-workspaces-ui-validation.json",
"sha256": "c95f933fafc7a6fa6e7e310d268536617ce54fa3b693eb0efb420ea2055d04d2",
"bytes": 1780
},
{
"path": ".qa/milestone-audit-verify-workspaces-ui.log",
"saved": "D:\\Ai-project\\REA逆向\\Agent-Workbench\\.qa\\milestone-first-candidate-3bd2cc18-321b-4a9e-800e-9ad645be72c2\\.qa\\milestone-audit-verify-workspaces-ui.log",
"sha256": "18fba3fd6ef6ed389fb4d74beaffb6c339a7240466c1b8e869c269840b78495a",
"bytes": 2608
},
{
"path": "docs/milestone-audit-python-ui-validation.json",
"saved": "D:\\Ai-project\\REA逆向\\Agent-Workbench\\.qa\\milestone-first-candidate-3bd2cc18-321b-4a9e-800e-9ad645be72c2\\docs\\milestone-audit-python-ui-validation.json",
"sha256": "966f446b85071212edb8d2e767c70ed1baed10e44972ce8a99eb97a7bf0ff1e1",
"bytes": 1374
},
{
"path": ".qa/milestone-audit-verify-python-ui.log",
"saved": "D:\\Ai-project\\REA逆向\\Agent-Workbench\\.qa\\milestone-first-candidate-3bd2cc18-321b-4a9e-800e-9ad645be72c2\\.qa\\milestone-audit-verify-python-ui.log",
"sha256": "755ebc650e88ef352b7cd2a0d4e6cce618c3c3d134b99a3935fc98c42bc7fec4",
"bytes": 2075
},
{
"path": "docs/milestone-audit-react-delivery-validation.json",
"saved": "D:\\Ai-project\\REA逆向\\Agent-Workbench\\.qa\\milestone-first-candidate-3bd2cc18-321b-4a9e-800e-9ad645be72c2\\docs\\milestone-audit-react-delivery-validation.json",
"sha256": "85cf65007ac5721bf259d8837da0aa2a471171e6bf4021d8062cdc4edf8c06ee",
"bytes": 2763
},
{
"path": ".qa/milestone-audit-verify-react-delivery.log",
"saved": "D:\\Ai-project\\REA逆向\\Agent-Workbench\\.qa\\milestone-first-candidate-3bd2cc18-321b-4a9e-800e-9ad645be72c2\\.qa\\milestone-audit-verify-react-delivery.log",
"sha256": "e6a396c5cf3e152862d4a842b013ce6d6520bc7bcc3f4728b216ad5a70f90ac3",
"bytes": 3426
},
{
"path": "docs/milestone-audit-research-ui-validation.json",
"saved": "D:\\Ai-project\\REA逆向\\Agent-Workbench\\.qa\\milestone-first-candidate-3bd2cc18-321b-4a9e-800e-9ad645be72c2\\docs\\milestone-audit-research-ui-validation.json",
"sha256": "554651c017eda8545cef34d30e5dfa561a32c7a32845f64643ba41f0ec45945a",
"bytes": 5419
},
{
"path": ".qa/milestone-audit-verify-research-ui.log",
"saved": "D:\\Ai-project\\REA逆向\\Agent-Workbench\\.qa\\milestone-first-candidate-3bd2cc18-321b-4a9e-800e-9ad645be72c2\\.qa\\milestone-audit-verify-research-ui.log",
"sha256": "5bc0378631934e45b6fb654d6c0ff148aa69f2f454ba3d871c833061139099ee",
"bytes": 1412
},
{
"path": "docs/milestone-audit-engines-ui-validation.json",
"saved": "D:\\Ai-project\\REA逆向\\Agent-Workbench\\.qa\\milestone-first-candidate-3bd2cc18-321b-4a9e-800e-9ad645be72c2\\docs\\milestone-audit-engines-ui-validation.json",
"sha256": "4e465d2360df5c36a9d8b5be6c5527495cc1e8a5bae24639b874890959a9efb8",
"bytes": 11332
},
{
"path": ".qa/milestone-audit-verify-engines-ui.log",
"saved": "D:\\Ai-project\\REA逆向\\Agent-Workbench\\.qa\\milestone-first-candidate-3bd2cc18-321b-4a9e-800e-9ad645be72c2\\.qa\\milestone-audit-verify-engines-ui.log",
"sha256": "5302b2109ab32773323203ad2b1276d6593573fc46168fcae9182b96fdd0280b",
"bytes": 1351
},
{
"path": "docs/milestone-audit-model-center-ui-validation.json",
"saved": "D:\\Ai-project\\REA逆向\\Agent-Workbench\\.qa\\milestone-first-candidate-3bd2cc18-321b-4a9e-800e-9ad645be72c2\\docs\\milestone-audit-model-center-ui-validation.json",
"sha256": "86579b41233e65c925699849247ac4c0e23e72b8cb7a926636b0a8581ebe7446",
"bytes": 780
},
{
"path": ".qa/milestone-audit-verify-model-center-ui.log",
"saved": "D:\\Ai-project\\REA逆向\\Agent-Workbench\\.qa\\milestone-first-candidate-3bd2cc18-321b-4a9e-800e-9ad645be72c2\\.qa\\milestone-audit-verify-model-center-ui.log",
"sha256": "0411af598092f0ce33d2408999cc94fcf99ec91e208ba6259bf5b28114815278",
"bytes": 593
},
{
"path": "docs/milestone-audit-project-memory-ui-validation.json",
"saved": "D:\\Ai-project\\REA逆向\\Agent-Workbench\\.qa\\milestone-first-candidate-3bd2cc18-321b-4a9e-800e-9ad645be72c2\\docs\\milestone-audit-project-memory-ui-validation.json",
"sha256": "6db3c3236fcd13bb49d22bd899ccad65ed88ae2bde02d8553a9b3ca58a23c0f2",
"bytes": 770
},
{
"path": ".qa/milestone-audit-verify-project-memory-ui.log",
"saved": "D:\\Ai-project\\REA逆向\\Agent-Workbench\\.qa\\milestone-first-candidate-3bd2cc18-321b-4a9e-800e-9ad645be72c2\\.qa\\milestone-audit-verify-project-memory-ui.log",
"sha256": "d31104015d899cdf37d1989fd3101597293165edb85b94e44689eb2b95b4fa53",
"bytes": 488
},
{
"path": "docs/milestone-audit-task-checkpoints-ui-validation.json",
"saved": "D:\\Ai-project\\REA逆向\\Agent-Workbench\\.qa\\milestone-first-candidate-3bd2cc18-321b-4a9e-800e-9ad645be72c2\\docs\\milestone-audit-task-checkpoints-ui-validation.json",
"sha256": "15cea3ed0910cdb30784e688b3a68538a7030c5c120601dcbfdac3711bdcc6b1",
"bytes": 610
},
{
"path": ".qa/milestone-audit-verify-task-checkpoints-ui.log",
"saved": "D:\\Ai-project\\REA逆向\\Agent-Workbench\\.qa\\milestone-first-candidate-3bd2cc18-321b-4a9e-800e-9ad645be72c2\\.qa\\milestone-audit-verify-task-checkpoints-ui.log",
"sha256": "3dcdbc2ad5cc33638d6807552712c59beedccd91e74322901d214e1506c394b0",
"bytes": 427
},
{
"path": "docs/milestone-audit-skills-ui-validation.json",
"saved": "D:\\Ai-project\\REA逆向\\Agent-Workbench\\.qa\\milestone-first-candidate-3bd2cc18-321b-4a9e-800e-9ad645be72c2\\docs\\milestone-audit-skills-ui-validation.json",
"sha256": "f318e8454e547a6e460beac80378c730b81456ebe87f0ce308c6dd34fe4c7792",
"bytes": 801
},
{
"path": ".qa/milestone-audit-verify-skills-ui.log",
"saved": "D:\\Ai-project\\REA逆向\\Agent-Workbench\\.qa\\milestone-first-candidate-3bd2cc18-321b-4a9e-800e-9ad645be72c2\\.qa\\milestone-audit-verify-skills-ui.log",
"sha256": "a6fea246baa09979534bcdeae07b93768ca69dc05a73daf81dec3319f384a2e1",
"bytes": 576
},
{
"path": "docs/milestone-audit-mcp-ui-validation.json",
"saved": "D:\\Ai-project\\REA逆向\\Agent-Workbench\\.qa\\milestone-first-candidate-3bd2cc18-321b-4a9e-800e-9ad645be72c2\\docs\\milestone-audit-mcp-ui-validation.json",
"sha256": "abe21da92912205e6dd7a29b303ce7b47cdc9adde0c0c70449994be3191db5db",
"bytes": 952
},
{
"path": ".qa/milestone-audit-verify-mcp-ui.log",
"saved": "D:\\Ai-project\\REA逆向\\Agent-Workbench\\.qa\\milestone-first-candidate-3bd2cc18-321b-4a9e-800e-9ad645be72c2\\.qa\\milestone-audit-verify-mcp-ui.log",
"sha256": "f1acbfb8cae75f01506fe108f25bba021267ad2a7c7f406fc4a8cbe950d928e2",
"bytes": 700
},
{
"path": "docs/milestone-audit-mcp-recovery-ui-validation.json",
"saved": "D:\\Ai-project\\REA逆向\\Agent-Workbench\\.qa\\milestone-first-candidate-3bd2cc18-321b-4a9e-800e-9ad645be72c2\\docs\\milestone-audit-mcp-recovery-ui-validation.json",
"sha256": "5fd28ed4bbc7740c01e9afff6fa099350c094bf3eb73213f263c7003d68ec8e1",
"bytes": 482
},
{
"path": ".qa/milestone-audit-verify-mcp-recovery-ui.log",
"saved": "D:\\Ai-project\\REA逆向\\Agent-Workbench\\.qa\\milestone-first-candidate-3bd2cc18-321b-4a9e-800e-9ad645be72c2\\.qa\\milestone-audit-verify-mcp-recovery-ui.log",
"sha256": "5c52cb33a03c5ce5937d808c58fa26ec8467fcd7fb0decda1307d8191490be03",
"bytes": 340
}
],
"directory": "D:\\Ai-project\\REA逆向\\Agent-Workbench\\.qa\\milestone-first-candidate-3bd2cc18-321b-4a9e-800e-9ad645be72c2",
"purpose": "Preserved initial candidate before hard-link boundary follow-up and earlier Windows alias rejection assertion; not final release evidence."
}
File diff suppressed because one or more lines are too long
File diff suppressed because it is too large Load Diff
@@ -0,0 +1,19 @@
{
"checkedAt": "2026-10-30T15:34:52.601Z",
"version": "1.12.1",
"passed": [
"工程索引可独立于任务使用,实际覆盖范围与损坏文件显示明确",
"真实AST查询打开原文件第181行并仅滚动代码容器,输入区和顶部保持可见",
"查询词保留;过期SHA不高亮旧位置,增量刷新后正确定位第184行",
"切换项目时延迟返回的旧索引不覆盖新项目结果",
"Python候选、字符串误报排除和局部导入文件候选保留真实证据标签",
"两处唯一锚点成为待审核提案,真实界面应用后BOM/CRLF和未选择原字节保留",
"390px索引查询、范围和完整指纹可读且无横向溢出,实际文件行定位通过",
"索引刷新失败保留归属和解释,旧结果禁用并可重试恢复"
],
"pageErrors": [],
"requestFailures": [],
"root": "D:\\Ai-project\\REA逆向\\Agent-Workbench\\.qa\\index-ui-f4472e28-4c3b-4c00-8715-904242dd0e02",
"primaryProjectId": "cb91ac22-dd97-4ece-99ad-98dd00baa327",
"otherProjectId": "f9ba3e33-0898-477e-a070-c1c48482bcef"
}
@@ -0,0 +1,16 @@
{
"success": true,
"assertions": [
"actual host crash with no result report exposes its original verified input and unknown status",
"390px original-input review clears the job and invocation gates while retaining unknown, no report and no automatic model replay"
],
"viewportWidths": [
1440,
390
],
"pageErrors": [],
"requestFailures": [],
"localFixtureModelRequests": 1,
"realModelRequests": 0,
"jobId": "1d497864-4449-41e1-bc4b-6e8bf82c2d78"
}
@@ -0,0 +1,24 @@
{
"success": true,
"assertions": [
"MCP registration stays disabled and shows the hashed local executable and Node entry",
"actual stdio tools/list schema is visible; only the reviewed echo tool is enabled",
"an enabled program whose source changes can be disabled immediately without executing the changed entry",
"approved external call interrupted after dispatch remains unknown and does not replay",
"TaskPanel exposes Task/Run source, original report, and human acknowledgement without marking unknown as success",
"390px MCP registry, long executable paths, controls, and operation history have no horizontal overflow"
],
"viewportWidths": [
1440,
390
],
"pageErrors": [],
"requestFailures": [],
"localFixtureModelRequests": 1,
"realModelRequests": 0,
"unknownJobId": "d8b9dea0-9abd-4026-93d4-3666b7f24bf0",
"screenshots": [
"desktop-mcp-registry.png",
"mobile-mcp-registry.png"
]
}
@@ -0,0 +1,22 @@
{
"url": "http://127.0.0.1:56571",
"viewportWidths": [
1440,
390
],
"assertions": [
"connection form stores only the volatile credential and renders configured status without exposing it",
"pool editor saves weighted members and binds selection for new Runs",
"schema-bound import preview, opt-in automatic sync, revision update and sync shutdown keep credentials hidden",
"local inspection shows real database and artifact checks and emits zero model requests",
"390px model-center modal and connection cards remain operable without horizontal page overflow"
],
"pageErrors": [],
"requestFailures": [],
"modelRequests": 0,
"screenshots": [
"desktop-operations.png",
"mobile-operations.png",
"mobile-connections.png"
]
}
@@ -0,0 +1,17 @@
{
"directory": "C:\\Users\\wangy\\AppData\\Local\\Temp\\awb-model-save-audit-h1AQcD",
"error": "fixture-disk-failure",
"before": {
"enabled": true,
"revision": 1
},
"memory": {
"enabled": false,
"revision": 2
},
"disk": {
"enabled": true,
"revision": 1
},
"modelRequests": 0
}
+22
View File
@@ -0,0 +1,22 @@
{
"auditReportVersion": 2,
"vulnerabilities": {},
"metadata": {
"vulnerabilities": {
"info": 0,
"low": 0,
"moderate": 0,
"high": 0,
"critical": 0,
"total": 0
},
"dependencies": {
"prod": 13,
"dev": 141,
"optional": 79,
"peer": 0,
"peerOptional": 0,
"total": 153
}
}
}
@@ -0,0 +1,34 @@
{
"root": "C:\\Users\\wangy\\AppData\\Local\\Temp\\awb-path-audit-BHnvFn",
"platform": "win32",
"results": [
{
"path": ".git/config",
"allowed": false,
"error": "该路径不在可读取的项目范围内。"
},
{
"path": "git-alias/config",
"allowed": true,
"protectedContentRead": true
},
{
"path": "credential-alias/config.txt",
"allowed": true,
"protectedContentRead": true
},
{
"path": "git-alias/new.txt",
"allowed": true,
"protectedContentRead": false
},
{
"path": ".git./config",
"allowed": true,
"protectedContentRead": false
},
{
"inventoryAliases": 0
}
]
}
@@ -0,0 +1,24 @@
{
"success": true,
"url": "http://127.0.0.1:58686",
"viewportWidths": [
1440,
390
],
"assertions": [
"source hashes remain visible and a changed file is explicitly expired with retained provenance",
"incomplete inventory marks existing facts unverified and blocks new saves",
"390px memory review stays within the viewport with the incomplete-scan warning visible",
"desktop and 390px request records expose the exact memory snapshot used by one owned local HTTP request"
],
"pageErrors": [],
"requestFailures": [],
"localFixtureModelRequests": 1,
"realModelRequests": 0,
"screenshots": [
"desktop-expired-memory.png",
"mobile-incomplete-memory.png",
"desktop-sent-memory.png",
"mobile-sent-memory.png"
]
}
@@ -0,0 +1,35 @@
{
"success": true,
"version": "1.12.1",
"at": "2026-10-30T15:35:57.032Z",
"browser": "156.0.4314.50",
"passed": [
"mixed projects retain npm defaults and explicitly seal the selected Python checks",
"the confirmation shows the actual interpreter and cancellation executes no checks",
"actual unittest failures and the separately passed compilation display their counted results",
"the log and original structured result are independently viewable and hash-bound",
"a reviewed three-file repair runs new checks while retaining the original failed job and protected files",
"390px checks, interpreter paths, original result dialog and composer remain operable without page overflow",
"Git confirmation and the Python execution report deliver only the three reviewed sources and preserve owner changes"
],
"pageErrors": [],
"requestFailures": [],
"checks": {
"before": {
"total": 6,
"passed": 1,
"failed": 5,
"skipped": 0,
"cancelled": 0
},
"after": {
"total": 6,
"passed": 6,
"failed": 0,
"skipped": 0,
"cancelled": 0
}
},
"commit": "3dba6ed970b8f9f792e5683d87c64e6ad728e711",
"scope": "Independent host and owned fixture. Real Python processes, reviewed API repairs and actual local Git. No configured model requests or production credentials."
}
@@ -0,0 +1,76 @@
{
"success": true,
"version": "1.12.1",
"at": "2026-10-30T15:36:17.052Z",
"browser": "156.0.4314.50",
"passed": [
"the sealed React TypeScript project builds but its real domain tests first expose the functional failures",
"a real browser separately observes that the broken pending filter wrongly shows all three tasks",
"the four-file reviewed repair passes all nine domain tests, actual TypeScript checking and the Vite production build",
"the production React app passes real filtering, search, single-task toggling and updated statistics",
"the same built React app remains operable at 390px without horizontal overflow",
"Workbench Git commits only the four reviewed files and binds the actual checks while preserving tests, configuration and owner changes"
],
"pageErrors": [],
"requestFailures": [],
"originalTests": {
"total": 9,
"passed": 2,
"failed": 7,
"cancelled": 0,
"skipped": 0
},
"repairedTests": {
"total": 9,
"passed": 9,
"failed": 0,
"cancelled": 0,
"skipped": 0
},
"beforePending": 3,
"checks": [
{
"name": "test",
"status": "passed",
"exitCode": 0,
"tests": {
"total": 9,
"passed": 9,
"failed": 0,
"cancelled": 0,
"skipped": 0
}
},
{
"name": "typecheck",
"status": "passed",
"exitCode": 0,
"tests": null
},
{
"name": "build",
"status": "passed",
"exitCode": 0,
"tests": null
}
],
"commit": "44d9f56dbf9d0ce3c9c2cd457dcfa006749d01e5",
"changes": [
"src/App.tsx",
"src/query.ts",
"src/summary.ts",
"src/transition.ts"
],
"protectedFiles": {
"package.json": "db3c6bfd79d1ebb1f5692727715e2b48d3f09e0314d9f1d455d335774f5e7db8",
"package-lock.json": "48c138a05408667bbd332000731598415c9ae281d5072daa18544025ad89f7ff",
"tsconfig.json": "9dcf064d622cb31ae94ef348655f4f7af84076241601148929232314ef1b0c8a",
"vite.config.mjs": "0ea84b4e2909b04e4c1cb1b92392a9f85bcb46605472763b15935d3498ba8ed3",
"index.html": "37e9fd3172422ecca3902378c567331363dc1277ac749dd9f6bdf74f04540787",
"test/domain.test.mjs": "3fbd155cfc4919f3d0570a7ee26ff5ff4243e3e94250f3d0dbebe93c2b925ec0",
"src/types.ts": "d2bbcac2cb175e7bcafe087915f02c3cf2e0ea4b65235d6231bbcd4c3a682246",
"src/main.tsx": "c9a5787f5544788757ab654e1dfd83b5bc51495144e17aa078c2ee2f4412a302",
"src/styles.css": "60bcb8f7f73f4328274096132a3d73b7ef1fc9e6fbf6a3a3477fb9d519f2c25c"
},
"scope": "Owned frozen React TypeScript project with offline-installed fixed dependencies, real npm/typecheck/build/browser and Workbench review/Git/report. API fixtures supply repairs; configured-model success is separate."
}
@@ -0,0 +1,138 @@
{
"success": true,
"version": "1.12.1",
"at": "2026-10-30T15:36:38.025Z",
"browser": "156.0.4314.50",
"passed": [
"independent desktop research entry detects the installed fixed REA version",
"UI registration seals the artifact identity in the current project and creates no model request",
"actual REA ASAR analysis renders the verified source chain and downloads the exact original Bundle",
"original field selection, paging and limitations retain exact bytes; Escape closes only the top dialog",
"raw responses and connection records are readable; repeat analysis reuses the verified original without dispatch",
"UI PE operation selection performs actual import and resource analysis with independently verified target hashes",
"changed artifact versions refuse dispatch, retain the stale record and require a separately registered version",
"390px research, long identities and original evidence dialogs fit the viewport and retain usable close controls",
"a real host restart preserves exact target/job identities and serves sealed Evidence without producer replay",
"a delayed evidence response cannot reopen a closed research panel or change another task",
"declared owned stdio fault fixture stops its root process and explicit retry adds a new record without changing history"
],
"pageErrors": [],
"requestFailures": [],
"expectedRequestFailures": [],
"root": "D:\\Ai-project\\REA逆向\\Agent-Workbench\\.qa\\research-ui-620b3e88-3c53-4077-b4e9-8a56541c5eed",
"actualRea": {
"version": "6.8.0",
"operations": [
"javascript",
"pe-imports",
"pe-resources"
],
"jobs": [
{
"id": "e75e35da-cf81-4f99-8264-a91b629c3a8d",
"status": "completed",
"evidenceId": "ev_2f04a55fd2f8b4d38ed986bfb50d4f2599f94b256a58371a496409498927b22e"
},
{
"id": "74b3f1a5-8738-432c-9e6c-9f41a5c97a22",
"status": "completed",
"evidenceId": "ev_2f04a55fd2f8b4d38ed986bfb50d4f2599f94b256a58371a496409498927b22e"
},
{
"id": "21a3902f-7cfe-4c43-9c03-808c59e8eb0f",
"status": "completed",
"evidenceId": "ev_92454ab0585d1ebbd2d89928228e1d2c1a14d8f67ace457ff54a59cadc8603ab"
},
{
"id": "114c223f-400e-4f2e-99ef-bc8fa36b88d6",
"status": "completed",
"evidenceId": "ev_d940b75287a3a711cef82cb8307b9673d7cb3ea27b5bc1d2759b3477580c4fb0"
},
{
"id": "48cd341e-dea6-4bb6-ab9a-bd5541e0fc9e",
"status": "stale"
},
{
"id": "a55450a8-378e-45ee-a2c1-3bb0af3832e3",
"status": "completed",
"evidenceId": "ev_ae449670ce6b68ae3b714a1d01777490144a22f5ee58eda5865ab0e4aba4bc1c"
}
],
"targetExecution": false,
"modelRequests": 0
},
"stopScope": "explicitly declared private stdio handshake fault fixture",
"screenshots": [
".qa/screenshots/milestone-audit-research-desktop.png",
".qa/screenshots/milestone-audit-research-mobile.png",
".qa/screenshots/milestone-audit-research-evidence-mobile.png",
".qa/screenshots/milestone-audit-research-stop-desktop.png"
],
"analysisRequests": [
{
"url": "http://127.0.0.1:52066/api/sessions/0288d57f-1388-4dbd-b750-6d15d2ab07b8/research/jobs",
"input": {
"targetId": "70838a70-475d-413d-b12e-fc0b262726c0",
"operation": "javascript",
"refresh": false
}
},
{
"url": "http://127.0.0.1:52066/api/sessions/0288d57f-1388-4dbd-b750-6d15d2ab07b8/research/jobs",
"input": {
"targetId": "70838a70-475d-413d-b12e-fc0b262726c0",
"operation": "javascript",
"refresh": false
}
},
{
"url": "http://127.0.0.1:52066/api/sessions/0288d57f-1388-4dbd-b750-6d15d2ab07b8/research/jobs",
"input": {
"targetId": "51c4f866-1141-42ce-9d12-baf3cbb49400",
"operation": "pe-imports",
"refresh": false
}
},
{
"url": "http://127.0.0.1:52066/api/sessions/0288d57f-1388-4dbd-b750-6d15d2ab07b8/research/jobs",
"input": {
"targetId": "51c4f866-1141-42ce-9d12-baf3cbb49400",
"operation": "pe-resources",
"refresh": false
}
},
{
"url": "http://127.0.0.1:52066/api/sessions/0288d57f-1388-4dbd-b750-6d15d2ab07b8/research/jobs",
"input": {
"targetId": "70838a70-475d-413d-b12e-fc0b262726c0",
"operation": "javascript",
"refresh": false
}
},
{
"url": "http://127.0.0.1:52066/api/sessions/0288d57f-1388-4dbd-b750-6d15d2ab07b8/research/jobs",
"input": {
"targetId": "61434dfe-2501-4f27-8204-c8b5599a8b29",
"operation": "javascript",
"refresh": false
}
},
{
"url": "http://127.0.0.1:60891/api/sessions/58255c34-c4f1-4416-9ebd-90c08ea059ca/research/jobs",
"input": {
"targetId": "56cd82b8-68ea-4a9f-9988-41b6e40dea7a",
"operation": "javascript",
"refresh": false
}
},
{
"url": "http://127.0.0.1:60891/api/sessions/58255c34-c4f1-4416-9ebd-90c08ea059ca/research/jobs",
"input": {
"targetId": "56cd82b8-68ea-4a9f-9988-41b6e40dea7a",
"operation": "javascript",
"refresh": true
}
}
],
"scope": "Independent ports/data and self-owned inert artifacts. No production service, configured-model request or real key read/copied."
}
@@ -0,0 +1,12 @@
{
"success": false,
"directory": "D:\\Ai-project\\REA逆向\\Agent-Workbench\\.qa\\milestone-security-ui-100d0d0e-d705-44a4-a6ae-0d1571541714",
"error": "AssertionError [ERR_ASSERTION]: Missing expected rejection.\n at process.processTicksAndRejections (node:internal/process/task_queues:104:5)\n at async file:///D:/Ai-project/REA%E9%80%86%E5%90%91/Agent-Workbench/.qa/milestone-security-ui.mjs:63:5",
"assertions": [
"real original 1.12.0 renders the workbench inside an independently hosted cross-origin iframe",
"an actual original 1.12.0 process upgrades in place with initialized ModelHub/Importer, unchanged PID, connection metadata and original session"
],
"pageErrors": [],
"expectedPolicyMessages": [],
"stderr": "Debugger listening on ws://127.0.0.1:9229/76b23bcc-01d1-47d4-b8cb-fddee85a2b15\r\nFor help, see: https://nodejs.org/learn/getting-started/debugging\r\nDebugger attached.\r\nDebugger ending on ws://127.0.0.1:9229/76b23bcc-01d1-47d4-b8cb-fddee85a2b15\r\nFor help, see: https://nodejs.org/learn/getting-started/debugging\r\n"
}
@@ -0,0 +1,83 @@
{
"success": true,
"version": "1.12.1",
"at": "2026-10-30T14:42:10.496Z",
"directory": "D:\\Ai-project\\REA逆向\\Agent-Workbench\\.qa\\milestone-security-ui-9f38a7ec-8649-4260-9c43-375425e6c8dd",
"oldReleaseId": "20261030124454517",
"oldHeaders": {
"cache-control": "no-store",
"connection": "keep-alive",
"content-type": "text/html; charset=utf-8",
"date": "Fri, 30 Oct 2026 14:42:06 GMT",
"keep-alive": "timeout=5",
"transfer-encoding": "chunked",
"x-content-type-options": "nosniff"
},
"newHeaders": {
"cache-control": "no-store",
"connection": "keep-alive",
"content-security-policy": "default-src 'self'; script-src 'self'; style-src 'self' 'unsafe-inline'; img-src 'self' data: blob:; connect-src 'self'; object-src 'none'; base-uri 'none'; frame-ancestors 'none'; form-action 'self'",
"content-type": "text/html; charset=utf-8",
"date": "Fri, 30 Oct 2026 14:42:07 GMT",
"keep-alive": "timeout=5",
"referrer-policy": "no-referrer",
"transfer-encoding": "chunked",
"x-content-type-options": "nosniff",
"x-frame-options": "DENY"
},
"assertions": [
"real original 1.12.0 renders the workbench inside an independently hosted cross-origin iframe",
"an actual original 1.12.0 process upgrades in place with initialized ModelHub/Importer, unchanged PID, connection metadata and original session",
"retained importer now rolls back an actual partially invalid batch and keeps the original preview usable after failure",
"both volatile fixture credentials still route correctly after the real process upgrade and are absent from persisted configuration",
"real Edge refuses the cross-origin iframe after the patch instead of exposing the approval interface",
"1280px workbench and model center remain usable under the new response policy",
"390px workbench and model center remain usable under the new response policy",
"policy permits PNG Blob image loading and blocks fetching/executing an independently hosted script"
],
"pageErrors": [],
"expectedPolicyMessages": [
"Framing 'http://127.0.0.1:56118/' violates the following Content Security Policy directive: \"frame-ancestors 'none'\". The request has been blocked.\n",
"Loading the script 'http://127.0.0.1:56121/forbidden.js' violates the following Content Security Policy directive: \"script-src 'self'\". Note that 'script-src-elem' was not explicitly set, so 'script-src' is used as a fallback. The action has been blocked."
],
"unexpectedRequests": [],
"pid": 36724,
"processExited": true,
"syntheticModelRequests": 2,
"realModelRequests": 0,
"files": [
{
"path": "D:\\Ai-project\\REA逆向\\Agent-Workbench\\.qa\\milestone-security-ui-9f38a7ec-8649-4260-9c43-375425e6c8dd\\buttons-1280.json",
"sha256": "ac701db35c9e4f392512f538843511c69ec57f6e8ddce660910b19c4c198d746"
},
{
"path": "D:\\Ai-project\\REA逆向\\Agent-Workbench\\.qa\\milestone-security-ui-9f38a7ec-8649-4260-9c43-375425e6c8dd\\buttons-390.json",
"sha256": "a8cd5c189a51fb9f990f93a5f86b66e8116edc67b33ee0096f7c1aa8b173f0a3"
},
{
"path": "D:\\Ai-project\\REA逆向\\Agent-Workbench\\.qa\\milestone-security-ui-9f38a7ec-8649-4260-9c43-375425e6c8dd\\model-center-1280.png",
"sha256": "e5110517113d81c9ff7a80294e1d9e0c28a487202ebb650a6e2acebd4c48f77f"
},
{
"path": "D:\\Ai-project\\REA逆向\\Agent-Workbench\\.qa\\milestone-security-ui-9f38a7ec-8649-4260-9c43-375425e6c8dd\\model-center-390.png",
"sha256": "e0fb690d0d92bb4da2565af70d9f507af05f701014ce7ca3c2e64f617439ca27"
},
{
"path": "D:\\Ai-project\\REA逆向\\Agent-Workbench\\.qa\\milestone-security-ui-9f38a7ec-8649-4260-9c43-375425e6c8dd\\new-blocked-iframe.png",
"sha256": "66ca379556d77ef1d88203acc3374643bda5896ca6846e79c61c6377769b61f4"
},
{
"path": "D:\\Ai-project\\REA逆向\\Agent-Workbench\\.qa\\milestone-security-ui-9f38a7ec-8649-4260-9c43-375425e6c8dd\\old-embedded.png",
"sha256": "001b1dfde3310ae50467e6091791a2769518784f4c063506c2d9e874b65e4c07"
},
{
"path": "D:\\Ai-project\\REA逆向\\Agent-Workbench\\.qa\\milestone-security-ui-9f38a7ec-8649-4260-9c43-375425e6c8dd\\old-frame.html",
"sha256": "af4c18ef5c86a7ef356580b2cf194822fe4ed55d6dd7ed950cacef3ddce77b64"
},
{
"path": "D:\\Ai-project\\REA逆向\\Agent-Workbench\\.qa\\milestone-security-ui-9f38a7ec-8649-4260-9c43-375425e6c8dd\\owned-import.json",
"sha256": "5efc64fd7e5c88e1c6c0c8572f9ef6b52a91643d3232a02f7152ad9c5fa442c5"
}
],
"scope": "Owned projects, credentials, independent ports and original immutable 1.12.0 process; no production service affected."
}
@@ -0,0 +1,119 @@
import fs from 'node:fs';
import path from 'node:path';
import http from 'node:http';
import assert from 'node:assert/strict';
import { spawn } from 'node:child_process';
import { randomUUID, createHash } from 'node:crypto';
import { chromium, expect } from 'playwright/test';
import { fixture, until, compatibleFixture, completion, workspace } from '../tests/helpers.mjs';
import { gitEnvironment } from '../src/core/git-process.mjs';
import { upgradeRunning } from '../scripts/upgrade-running.mjs';
const directory = path.resolve('.qa/milestone-security-ui-' + randomUUID()); fs.mkdirSync(directory);
const original = JSON.parse(fs.readFileSync('docs/model-center-release-identity.json', 'utf8'));
const oldRoot = original.release.portableRoot;
const disposers = [], t = { after: action => disposers.push(action) }, f = fixture(t); f.store.close();
const assertions = [], pageErrors = [], expectedPolicyMessages = [], unexpectedRequests = [];
const record = value => { assertions.push(value); console.log('PASS ' + value); };
const fakeModel = await compatibleFixture(t, () => completion([], 'owned model probe response'));
const child = spawn(process.execPath, [path.join(oldRoot, 'dist-server/server.mjs')], { cwd: oldRoot, env: gitEnvironment({
AGENT_WORKBENCH_APP_DIR: oldRoot, AGENT_WORKBENCH_DATA_DIR: f.data, AGENT_WORKBENCH_PORT: '0' }), windowsHide: true, stdio: ['ignore', 'pipe', 'pipe'] });
let url = '', stdout = '', stderr = '', browser, external;
child.stdout.on('data', bytes => { stdout += bytes.toString(); for (const line of stdout.split('\n')) { try { const item = JSON.parse(line); if (item.ready) url = item.url; } catch {} } });
child.stderr.on('data', bytes => { stderr += bytes.toString(); });
let bootstrap;
async function api(route, input) {
const response = await fetch(url + '/api' + route, { method: input === undefined ? 'GET' : 'POST', headers: {
'Content-Type': 'application/json', 'X-Workbench-Token': bootstrap.token, Connection: 'close' }, ...(input === undefined ? {} : { body: JSON.stringify(input) }) });
const value = await response.json(); if (!response.ok) throw Object.assign(new Error(value.error), { status: response.status }); return value;
}
try {
await until(() => url || child.exitCode !== null); assert.ok(url, stderr); assert.notEqual(new URL(url).port, '4318');
bootstrap = await (await fetch(url + '/api/bootstrap')).json(); assert.equal(bootstrap.version, '1.12.0'); assert.equal(bootstrap.config.hasKey, false);
const oldHeaders = Object.fromEntries((await fetch(url)).headers); assert.equal(oldHeaders['x-frame-options'], undefined);
external = http.createServer((request, response) => {
if (request.url === '/forbidden.js') { unexpectedRequests.push('external script fetched'); response.writeHead(200, { 'Content-Type': 'text/javascript' }); response.end('window.ownedExternalScriptExecuted=true;'); }
else { response.writeHead(200, { 'Content-Type': 'text/html' }); response.end('<!doctype html><title>Owned embedding fixture</title><h1>External fixture</h1><iframe id="workbench" src="' + url + '" style="width:1100px;height:760px"></iframe>'); }
});
await new Promise(resolve => external.listen(0, '127.0.0.1', resolve)); const externalUrl = 'http://127.0.0.1:' + external.address().port;
browser = await chromium.launch({ channel: 'msedge', headless: true });
const page = await browser.newPage({ viewport: { width: 1280, height: 1000 } });
page.on('pageerror', error => pageErrors.push(error.message));
page.on('console', message => { if (message.type() === 'error' && /frame|Content Security Policy|script-src/i.test(message.text())) expectedPolicyMessages.push(message.text()); });
await page.goto(externalUrl); await page.waitForLoadState('networkidle');
await expect(page.frameLocator('#workbench').getByRole('button', { name: '打开模型中心', exact: true })).toBeVisible();
fs.writeFileSync(path.join(directory, 'old-frame.html'), await page.content()); await page.screenshot({ path: path.join(directory, 'old-embedded.png') });
record('real original 1.12.0 renders the workbench inside an independently hosted cross-origin iframe');
await api('/config', { mode: 'compatible', model: 'owned-legacy', baseUrl: fakeModel.baseUrl, apiKey: 'owned-legacy-upgrade-marker' });
const connection = await api('/model-center/connections', { name: 'Retained owned connection', baseUrl: fakeModel.baseUrl, model: 'owned-reserve', apiKey: 'owned-connection-upgrade-marker' });
const importFile = path.join(directory, 'owned-import.json');
fs.writeFileSync(importFile, JSON.stringify({ schema: 1, connections: [
{ name: 'First valid imported entry', baseUrl: fakeModel.baseUrl, model: 'owned-first', apiKey: 'owned-import-marker' },
{ name: 'Second invalid imported entry', baseUrl: fakeModel.baseUrl, model: 'owned-invalid', apiKey: 'x'.repeat(16001) },
] }));
const preview = await api('/model-center/import-preview', { source: 'workbench-json', path: importFile });
const beforeCenter = await api('/model-center'), beforeSession = await api('/sessions/' + f.session.id), beforeConfig = (await api('/bootstrap')).config;
const upgraded = await upgradeRunning({ url, pid: child.pid, currentRoot: oldRoot, newRoot: workspace });
assert.equal(upgraded.verified, true); assert.equal(upgraded.version, '1.12.1'); assert.equal(upgraded.pid, child.pid); assert.equal(upgraded.hasKey, true);
bootstrap = await api('/bootstrap'); assert.deepEqual(bootstrap.config, beforeConfig);
assert.deepEqual(await api('/model-center'), beforeCenter); assert.deepEqual(await api('/sessions/' + f.session.id), beforeSession);
record('an actual original 1.12.0 process upgrades in place with initialized ModelHub/Importer, unchanged PID, connection metadata and original session');
for (let attempt = 0; attempt < 2; attempt++) {
await assert.rejects(api('/model-center/import-apply', { previewId: preview.id }), error => error.status === 400 && !/预览已过期/.test(error.message));
assert.deepEqual(await api('/model-center'), beforeCenter);
}
record('retained importer now rolls back an actual partially invalid batch and keeps the original preview usable after failure');
assert.equal((await api('/model-center/probe', { id: connection.id })).ok, true);
assert.equal(fakeModel.requests.at(-1).headers.authorization, 'Bearer owned-connection-upgrade-marker');
await api('/config/models', {}); assert.equal(fakeModel.requests.at(-1).headers.authorization, 'Bearer owned-legacy-upgrade-marker');
assert.ok(!fs.readFileSync(path.join(f.data, 'config.json'), 'utf8').includes('owned-legacy-upgrade-marker'));
record('both volatile fixture credentials still route correctly after the real process upgrade and are absent from persisted configuration');
const newHeaders = Object.fromEntries((await fetch(url)).headers); assert.equal(newHeaders['x-frame-options'], 'DENY');
assert.match(newHeaders['content-security-policy'], /frame-ancestors 'none'/); assert.equal(newHeaders['referrer-policy'], 'no-referrer');
await page.goto(externalUrl); await page.waitForLoadState('networkidle');
assert.equal(await page.frameLocator('#workbench').locator('#root').count(), 0);
assert.ok(expectedPolicyMessages.some(value => /frame-ancestors|frame/i.test(value)));
await page.screenshot({ path: path.join(directory, 'new-blocked-iframe.png') });
record('real Edge refuses the cross-origin iframe after the patch instead of exposing the approval interface');
await page.addInitScript(({ projectId, sessionId }) => { localStorage.setItem('awb-project', projectId); localStorage.setItem('awb-session', sessionId); }, { projectId: f.project.id, sessionId: f.session.id });
for (const width of [1280, 390]) {
await page.setViewportSize({ width, height: 1000 }); await page.goto(url); await page.waitForLoadState('networkidle');
fs.writeFileSync(path.join(directory, 'buttons-' + width + '.json'), JSON.stringify(await page.locator('button').evaluateAll(nodes => nodes.map(node => ({ text: node.textContent, label: node.getAttribute('aria-label'), title: node.title }))), null, 2));
await page.getByRole('button', { name: '打开模型中心', exact: true }).click();
const dialog = page.getByRole('dialog', { name: '模型中心', exact: true }); await expect(dialog).toBeVisible();
await expect(dialog).toContainText('Retained owned connection');
const bounds = await dialog.boundingBox(); assert.ok(bounds.x >= 0 && bounds.x + bounds.width <= width + 1);
await page.screenshot({ path: path.join(directory, 'model-center-' + width + '.png') }); await page.keyboard.press('Escape');
assert.equal(await page.evaluate(() => document.documentElement.scrollWidth <= innerWidth), true);
record(width + 'px workbench and model center remain usable under the new response policy');
}
const blobImage = await page.evaluate(async () => {
const bytes = Uint8Array.from(atob('iVBORw0KGgoAAAANSUhEUgAAAAEAAAABCAQAAAC1HAwCAAAAC0lEQVR42mP8/x8AAwMCAO+/lWQAAAAASUVORK5CYII='), value => value.charCodeAt(0));
const url = URL.createObjectURL(new Blob([bytes], { type: 'image/png' })), image = document.createElement('img');
try { const loaded = await new Promise(resolve => { image.onload = () => resolve(image.naturalWidth); image.onerror = () => resolve(0); image.src = url; document.body.append(image); }); return loaded; }
finally { image.remove(); URL.revokeObjectURL(url); }
});
assert.equal(blobImage, 1);
await assert.rejects(page.addScriptTag({ url: externalUrl + '/forbidden.js' }));
assert.equal(await page.evaluate(() => window.ownedExternalScriptExecuted === true), false); assert.deepEqual(unexpectedRequests, []);
record('policy permits PNG Blob image loading and blocks fetching/executing an independently hosted script');
assert.deepEqual(pageErrors, []); await browser.close(); browser = null;
await api('/shutdown', {}); await until(() => child.exitCode !== null); assert.equal(child.exitCode, 0);
await assert.rejects(fetch('http://127.0.0.1:9229/json/list', { signal: AbortSignal.timeout(500) }));
const hash = filename => createHash('sha256').update(fs.readFileSync(filename)).digest('hex');
const files = fs.readdirSync(directory).map(name => ({ path: path.join(directory, name), sha256: hash(path.join(directory, name)) }));
const result = { success: true, version: '1.12.1', at: new Date().toISOString(), directory, oldReleaseId: original.releaseId,
oldHeaders, newHeaders, assertions, pageErrors, expectedPolicyMessages, unexpectedRequests, pid: child.pid, processExited: true,
syntheticModelRequests: fakeModel.requests.length, realModelRequests: 0, files, scope: 'Owned projects, credentials, independent ports and original immutable 1.12.0 process; no production service affected.' };
fs.writeFileSync(path.join(directory, 'result.json'), JSON.stringify(result, null, 2) + '\n');
fs.writeFileSync('.qa/milestone-security-ui-result.json', JSON.stringify(result, null, 2) + '\n'); console.log('EVIDENCE ' + directory);
} catch (error) {
fs.writeFileSync(path.join(directory, 'failure.json'), JSON.stringify({ success: false, directory, error: error.stack, assertions, pageErrors, expectedPolicyMessages, stderr }, null, 2)); throw error;
} finally {
await browser?.close(); if (external) await new Promise(resolve => external.close(resolve));
if (child.exitCode === null && child.signalCode === null) { try { await api('/shutdown', {}); await until(() => child.exitCode !== null, 5000); } catch { child.kill(); } }
for (const dispose of disposers.reverse()) await dispose();
}
@@ -0,0 +1,21 @@
{
"observedAt": "2026-10-30T15:34:04.6420682Z",
"version": "1.12.0",
"pid": 23172,
"registeredPid": 23172,
"appRoot": "D:\\Ai-project\\REA逆向\\Agent-Workbench\\releases\\Agent-Workbench-1.12.0-win-x64-20261030124454517",
"runtimeRoot": "D:\\Ai-project\\REA逆向\\Agent-Workbench\\releases\\Agent-Workbench-1.12.0-win-x64-20261030124454517",
"dataRoot": "C:\\Users\\wangy\\AppData\\Local\\AgentWorkbench",
"processExecutable": "D:\\Ai-project\\REA逆向\\Agent-Workbench\\releases\\Agent-Workbench-1.12.0-win-x64-20261030124454517\\runtime\\node.exe",
"processCreatedAt": "2026-10-30T14:33:10.3641640Z",
"hasKey": true,
"mode": "compatible",
"model": "gpt-6.1-sol",
"visibleProjects": 16,
"visibleSessions": 20,
"busySessions": 0,
"historicalReleasePid": 19084,
"historicalPidIsCurrent": false,
"realModelRequests": 0,
"scope": "Read-only current host identity; public hasKey only. Upgrade must preserve this current process and freshly snapshot every persisted row/export, rather than use historical release counts."
}
@@ -0,0 +1,22 @@
{
"success": true,
"viewportWidths": [
1440,
390
],
"assertions": [
"desktop registers a disabled skill, reviews original UTF-8 instructions and explicitly enables it with no model calls",
"source edit blocks activation until a new reviewed revision is registered; original revision remains available",
"390px review and activation retain readable paths, revision history and visible source state without horizontal overflow",
"actual local HTTP task reads precisely revision 2 and exposes its SHA-bound sent catalog in task evidence"
],
"pageErrors": [],
"requestFailures": [],
"localFixtureModelRequests": 2,
"realModelRequests": 0,
"screenshots": [
"desktop-skill-source.png",
"mobile-skill-source.png",
"desktop-sent-skill-catalog.png"
]
}
@@ -0,0 +1,35 @@
{
"success": true,
"version": "1.12.1",
"startedAt": "2026-10-30T14:33:49.472Z",
"finishedAt": "2026-10-30T14:35:40.420Z",
"directory": "D:\\Ai-project\\REA逆向\\Agent-Workbench\\.qa\\milestone-targeted-673244ae-32ef-46d2-9719-63d73d92598d",
"files": [
"tests/files.test.mjs",
"tests/patch.test.mjs",
"tests/code-tools.test.mjs",
"tests/model-center.test.mjs",
"tests/upgrade-model-center.test.mjs",
"tests/api.test.mjs",
"tests/upgrade.test.mjs",
"tests/process-environment.test.mjs",
"tests/checks.test.mjs",
"tests/python-checks.test.mjs",
"tests/runtime.test.mjs"
],
"code": 0,
"tests": {
"tests": 80,
"pass": 80,
"fail": 0,
"cancelled": 0,
"skipped": 0,
"todo": 0
},
"log": {
"path": "D:\\Ai-project\\REA逆向\\Agent-Workbench\\.qa\\milestone-targeted-673244ae-32ef-46d2-9719-63d73d92598d\\tests.tap",
"bytes": 20775,
"sha256": "11fa9751969c13d1c1544c4b9eda460ecfbf19de13319d3def71e9b5043cef74"
},
"realModelRequests": 0
}
+490
View File
@@ -0,0 +1,490 @@
TAP version 13
# Subtest: the bundled local service enforces host, origin, token and JSON object contracts
ok 1 - the bundled local service enforces host, origin, token and JSON object contracts
---
duration_ms: 1140.4923
type: 'test'
...
# Subtest: an in-flight connection probe prevents configuration changes and another probe until its actual result returns
ok 2 - an in-flight connection probe prevents configuration changes and another probe until its actual result returns
---
duration_ms: 1150.652
type: 'test'
...
# Subtest: API review applies verified content, detects conflicts, exports sessions and keeps keys in memory
ok 3 - API review applies verified content, detects conflicts, exports sessions and keeps keys in memory
---
duration_ms: 2027.8642
type: 'test'
...
# Subtest: a real bundle restart preserves proposals, interrupts old approvals and requests fresh consent on resume
ok 4 - a real bundle restart preserves proposals, interrupts old approvals and requests fresh consent on resume
---
duration_ms: 7285.0082
type: 'test'
...
# Subtest: runtime diagnostics and model-response validation use actual records and preserve the saved connection
ok 5 - runtime diagnostics and model-response validation use actual records and preserve the saved connection
---
duration_ms: 1269.4122
type: 'test'
...
# Subtest: persistent event replay is read-only and keeps run/artifact identity across reconnects
ok 6 - persistent event replay is read-only and keeps run/artifact identity across reconnects
---
duration_ms: 6534.977
type: 'test'
...
# Subtest: engineering inspection preserves a dirty Git baseline and treats project rules as data
ok 7 - engineering inspection preserves a dirty Git baseline and treats project rules as data
---
duration_ms: 2489.6531
type: 'test'
...
# Subtest: real npm tests and build produce counted results, log artifacts and persistent verification
ok 8 - real npm tests and build produce counted results, log artifacts and persistent verification
---
duration_ms: 2266.4572
type: 'test'
...
# Subtest: an actual failing test retains the nonzero exit code and failed count
ok 9 - an actual failing test retains the nonzero exit code and failed count
---
duration_ms: 2121.7752
type: 'test'
...
# Subtest: zero tests, skipped tests and unrecognized stdout cannot become verification passes
ok 10 - zero tests, skipped tests and unrecognized stdout cannot become verification passes
---
duration_ms: 7113.8357
type: 'test'
...
# Subtest: a successful build with no selected tests does not verify the task
ok 11 - a successful build with no selected tests does not verify the task
---
duration_ms: 1743.1807
type: 'test'
...
# Subtest: stale check plans and forged script names execute no project commands
ok 12 - stale check plans and forged script names execute no project commands
---
duration_ms: 1443.4315
type: 'test'
...
# Subtest: applying a later proposal invalidates the prior passed engineering snapshot
ok 13 - applying a later proposal invalidates the prior passed engineering snapshot
---
duration_ms: 2744.2162
type: 'test'
...
# Subtest: stopping an actual check terminates its child process and blocks concurrent jobs
ok 14 - stopping an actual check terminates its child process and blocks concurrent jobs
---
duration_ms: 1834.6345
type: 'test'
...
# Subtest: check timeout records a failure with its actual termination result
ok 15 - check timeout records a failure with its actual termination result
---
duration_ms: 2097.9352
type: 'test'
...
# Subtest: source edits during a successful check make the result stale
ok 16 - source edits during a successful check make the result stale
---
duration_ms: 3248.9389
type: 'test'
...
# Subtest: bounded large logs retain the test footer and readable artifact instead of inventing counts
ok 17 - bounded large logs retain the test footer and readable artifact instead of inventing counts
---
duration_ms: 2239.0116
type: 'test'
...
# Subtest: model-selected checks require a run-bound decision and denial creates no check job
ok 18 - model-selected checks require a run-bound decision and denial creates no check job
---
duration_ms: 3607.9929
type: 'test'
...
# Subtest: check APIs expose owned plans and jobs and preserve the authoritative results across restart
ok 19 - check APIs expose owned plans and jobs and preserve the authoritative results across restart
---
duration_ms: 4102.6016
type: 'test'
...
# Subtest: reviewed changes continue the original task without reopening a generic completed run
ok 20 - reviewed changes continue the original task without reopening a generic completed run
---
duration_ms: 3861.4867
type: 'test'
...
# Subtest: concurrent review-continuation API requests accept one run and retain the goal after restart
ok 21 - concurrent review-continuation API requests accept one run and retain the goal after restart
---
duration_ms: 7416.1113
type: 'test'
...
# Subtest: the model reads owned failure logs without running the check again
ok 22 - the model reads owned failure logs without running the check again
---
duration_ms: 4454.8987
type: 'test'
...
# Subtest: process exit at check queued preserves results and never replays commands
ok 23 - process exit at check queued preserves results and never replays commands
---
duration_ms: 1548.6213
type: 'test'
...
# Subtest: process exit at check result_committed preserves results and never replays commands
ok 24 - process exit at check result_committed preserves results and never replays commands
---
duration_ms: 2832.095
type: 'test'
...
# Subtest: model code tools keep Task/Run ownership, index evidence and exact proposals across storage recovery
ok 25 - model code tools keep Task/Run ownership, index evidence and exact proposals across storage recovery
---
duration_ms: 4625.6807
type: 'test'
...
# Subtest: bundled code APIs isolate project results, validate queries and create review-only patches with conflicts
ok 26 - bundled code APIs isolate project results, validate queries and create review-only patches with conflicts
---
duration_ms: 2189.429
type: 'test'
...
# Subtest: code tool arguments reject malformed ownership-independent inputs before side effects
ok 27 - code tool arguments reject malformed ownership-independent inputs before side effects
---
duration_ms: 0.4268
type: 'test'
...
# Subtest: proposals preserve project bytes until review; apply backs up the exact original bytes
ok 28 - proposals preserve project bytes until review; apply backs up the exact original bytes
---
duration_ms: 861.7613
type: 'test'
...
# Subtest: new nested files are created only on apply and unchanged content creates no proposal
ok 29 - new nested files are created only on apply and unchanged content creates no proposal
---
duration_ms: 1087.478
type: 'test'
...
# Subtest: stale reads and changed baselines are rejected without overwriting external edits
ok 30 - stale reads and changed baselines are rejected without overwriting external edits
---
duration_ms: 746.7513
type: 'test'
...
# Subtest: file paths reject traversal, absolute paths, credentials and outside junctions
ok 31 - file paths reject traversal, absolute paths, credentials and outside junctions
---
duration_ms: 699.1052
type: 'test'
...
# Subtest: read and search expose bounded text scopes and reject non UTF-8 / oversized files
ok 32 - read and search expose bounded text scopes and reject non UTF-8 / oversized files
---
duration_ms: 812.157
type: 'test'
...
# Subtest: project-local junctions cannot expose or create protected Git and credential files
ok 33 - project-local junctions cannot expose or create protected Git and credential files
---
duration_ms: 804.7499
type: 'test'
...
# Subtest: Windows trailing-dot and trailing-space path aliases cannot bypass protected paths
ok 34 - Windows trailing-dot and trailing-space path aliases cannot bypass protected paths
---
duration_ms: 796.5457
type: 'test'
...
# Subtest: model registry keeps original credentials/endpoint intact, rejects stale edits, and forgets only volatile keys on restart
ok 35 - model registry keeps original credentials/endpoint intact, rejects stale edits, and forgets only volatile keys on restart
---
duration_ms: 1482.7296
type: 'test'
...
# Subtest: config import previews exact revisions, keeps secrets volatile, syncs only when enabled and preserves manual edits on conflicts
ok 36 - config import previews exact revisions, keeps secrets volatile, syncs only when enabled and preserves manual edits on conflicts
---
duration_ms: 1057.5669
type: 'test'
...
# Subtest: real HTTP pool fails over 429 to another model and never repeats completed tools; attempts seal actual request bytes and Run identity
ok 37 - real HTTP pool fails over 429 to another model and never repeats completed tools; attempts seal actual request bytes and Run identity
---
duration_ms: 2746.5494
type: 'test'
...
# Subtest: pool rotation and weights select members on new Runs; disabled and missing credentials are excluded without requests
ok 38 - pool rotation and weights select members on new Runs; disabled and missing credentials are excluded without requests
---
duration_ms: 6161.3018
type: 'test'
...
# Subtest: 401 blocks the credential revision; updating the key restores availability without leaking error echoes
ok 39 - 401 blocks the credential revision; updating the key restores availability without leaking error echoes
---
duration_ms: 2059.4583
type: 'test'
...
# Subtest: queued request cancellation and in-flight disable release only owned leases and do not dispatch another HTTP request
ok 40 - queued request cancellation and in-flight disable release only owned leases and do not dispatch another HTTP request
---
duration_ms: 1265.7207
type: 'test'
...
# Subtest: compiled model-center HTTP API persists public profiles, blocks forged origins, and local operations never invoke a model
ok 41 - compiled model-center HTTP API persists public profiles, blocks forged origins, and local operations never invoke a model
---
duration_ms: 1422.0311
type: 'test'
...
# Subtest: model registry disk failures restore routing, pool profiles and credential references without partial updates
ok 42 - model registry disk failures restore routing, pool profiles and credential references without partial updates
---
duration_ms: 1023.5952
type: 'test'
...
# Subtest: model selection and saved configuration commit together and failures preserve the original mode and endpoint
ok 43 - model selection and saved configuration commit together and failures preserve the original mode and endpoint
---
duration_ms: 1023.7873
type: 'test'
...
# Subtest: configuration imports roll back the entire batch and retain one usable preview after validation or commit failure
ok 44 - configuration imports roll back the entire batch and retain one usable preview after validation or commit failure
---
duration_ms: 683.1714
type: 'test'
...
# Subtest: exact patches preserve BOM/CRLF, unicode and all unselected bytes and back up the original on apply
ok 45 - exact patches preserve BOM/CRLF, unicode and all unselected bytes and back up the original on apply
---
duration_ms: 855.5786
type: 'test'
...
# Subtest: several patches are matched against one original snapshot in any input order and keep LF
ok 46 - several patches are matched against one original snapshot in any input order and keep LF
---
duration_ms: 914.334
type: 'test'
...
# Subtest: mixed line endings require exact anchors and preserve untouched formatting
ok 47 - mixed line endings require exact anchors and preserve untouched formatting
---
duration_ms: 802.8502
type: 'test'
...
# Subtest: ambiguous, overlapping, missing and malformed patch anchors never create a proposal or write
ok 48 - ambiguous, overlapping, missing and malformed patch anchors never create a proposal or write
---
duration_ms: 973.8542
type: 'test'
...
# Subtest: stale original hashes and changes before apply preserve owner edits and mark conflicts
ok 49 - stale original hashes and changes before apply preserve owner edits and mark conflicts
---
duration_ms: 1050.0935
type: 'test'
...
# Subtest: no-op BOM patches/full content produce no change and replacements cannot split Unicode bytes
ok 50 - no-op BOM patches/full content produce no change and replacements cannot split Unicode bytes
---
duration_ms: 675.0859
type: 'test'
...
# Subtest: an actual shell command keeps Node and UTF-8 output while omitting host-only credentials
ok 51 - an actual shell command keeps Node and UTF-8 output while omitting host-only credentials
---
duration_ms: 1124.9693
type: 'test'
...
# Subtest: real npm tests and build do not inherit arbitrary host credentials or Node injection settings
ok 52 - real npm tests and build do not inherit arbitrary host credentials or Node injection settings
---
duration_ms: 2274.3652
type: 'test'
...
# Subtest: real Python tests and compilation keep their sealed inputs without host credentials
ok 53 - real Python tests and compilation keep their sealed inputs without host credentials
---
duration_ms: 2404.7429
type: 'test'
...
# Subtest: Python static inspection does not import project code or probe a project interpreter before approval
ok 54 - Python static inspection does not import project code or probe a project interpreter before approval
---
duration_ms: 1216.2853
type: 'test'
...
# Subtest: a real three-module Python failure, reviewed repairs, tests and compilation retain exact structured evidence
ok 55 - a real three-module Python failure, reviewed repairs, tests and compilation retain exact structured evidence
---
duration_ms: 4902.4376
type: 'test'
...
# Subtest: unittest counts repeated execution ordinals and subtest skips rather than deduplicating display names
ok 56 - unittest counts repeated execution ordinals and subtest skips rather than deduplicating display names
---
duration_ms: 3427.6883
type: 'test'
...
# Subtest: unittest empty, skipped, expected failure and class initialization errors never verify a task
ok 57 - unittest empty, skipped, expected failure and class initialization errors never verify a task
---
duration_ms: 8458.5265
type: 'test'
...
# Subtest: stdout claims cannot replace actual unittest results and an early exit cannot forge a successful collector
ok 58 - stdout claims cannot replace actual unittest results and an early exit cannot forge a successful collector
---
duration_ms: 3386.228
type: 'test'
...
# Subtest: compilation reads actual Python source bytes without executing top-level project code
ok 59 - compilation reads actual Python source bytes without executing top-level project code
---
duration_ms: 3085.9532
type: 'test'
...
# Subtest: Node and Python mixed projects retain the existing npm default and require an explicit Python selection
ok 60 - Node and Python mixed projects retain the existing npm default and require an explicit Python selection
---
duration_ms: 2593.428
type: 'test'
...
# Subtest: changed project interpreters and forged Python plan bindings execute no check
ok 61 - changed project interpreters and forged Python plan bindings execute no check
---
duration_ms: 1784.3921
type: 'test'
...
# Subtest: malformed TOML and invalid Python test paths are reported without expanding npm check scope
ok 62 - malformed TOML and invalid Python test paths are reported without expanding npm check scope
---
duration_ms: 1434.1686
type: 'test'
...
# Subtest: stopping a real Python check terminates its child and keeps the completed result as cancelled
ok 63 - stopping a real Python check terminates its child and keeps the completed result as cancelled
---
duration_ms: 2354.4609
type: 'test'
...
# Subtest: real pytest pass, failure, skip, collection errors and teardown errors use collected case outcomes
ok 64 - real pytest pass, failure, skip, collection errors and teardown errors use collected case outcomes
---
duration_ms: 13021.3619
type: 'test'
...
# Subtest: a real project virtual environment is detected statically and its actual version is recorded only during the approved check
ok 65 - a real project virtual environment is detected statically and its actual version is recorded only during the approved check
---
duration_ms: 5243.031
type: 'test'
...
# Subtest: a missing interpreter and repeated pytest item objects cannot create a false verification pass
ok 66 - a missing interpreter and repeated pytest item objects cannot create a false verification pass
---
duration_ms: 2624.9358
type: 'test'
...
# Subtest: a stop from the event loop interrupts immediate model replies before later tools
ok 67 - a stop from the event loop interrupts immediate model replies before later tools
---
duration_ms: 1907.2383
type: 'test'
...
# Subtest: a stop after the first tool preserves a complete interrupted multi-call protocol group
ok 68 - a stop after the first tool preserves a complete interrupted multi-call protocol group
---
duration_ms: 1833.902
type: 'test'
...
# Subtest: the compatible HTTP loop preserves call IDs, validates errors and waits before executing commands
ok 69 - the compatible HTTP loop preserves call IDs, validates errors and waits before executing commands
---
duration_ms: 5221.645
type: 'test'
...
# Subtest: denial prevents command execution and produces a paired denied result
ok 70 - denial prevents command execution and produces a paired denied result
---
duration_ms: 1901.2261
type: 'test'
...
# Subtest: stopping during approval interrupts pending calls without executing or replaying them
ok 71 - stopping during approval interrupts pending calls without executing or replaying them
---
duration_ms: 1527.4161
type: 'test'
...
# Subtest: stopping a running fixture command terminates its own process tree and retains uncertain state
ok 72 - stopping a running fixture command terminates its own process tree and retains uncertain state
---
duration_ms: 2435.5711
type: 'test'
...
# Subtest: command output decodes split UTF-8 and excludes the workbench key from child environment
ok 73 - command output decodes split UTF-8 and excludes the workbench key from child environment
---
duration_ms: 1223.7402
type: 'test'
...
# Subtest: model-list testing uses a tentative key without changing the saved connection and redacts service errors
ok 74 - model-list testing uses a tentative key without changing the saved connection and redacts service errors
---
duration_ms: 6.7258
type: 'test'
...
# Subtest: truncated or repeated model calls fail before execution; tool input validation rejects unknown capabilities
ok 75 - truncated or repeated model calls fail before execution; tool input validation rejects unknown capabilities
---
duration_ms: 2285.6172
type: 'test'
...
# Subtest: request projection keeps exact tool pairs and complete recent user instructions while preserving disk history
ok 76 - request projection keeps exact tool pairs and complete recent user instructions while preserving disk history
---
duration_ms: 1324.0526
type: 'test'
...
# Subtest: hot upgrade refreshes retained model and importer methods while preserving credential and routing objects
ok 77 - hot upgrade refreshes retained model and importer methods while preserving credential and routing objects
---
duration_ms: 745.7607
type: 'test'
...
# Subtest: hot upgrade refuses active probes and restores retained model methods if installation fails
ok 78 - hot upgrade refuses active probes and restores retained model methods if installation fails
---
duration_ms: 694.9354
type: 'test'
...
# Subtest: a directly started compiled host upgrades twice with the same PID, sessions, applied bytes and volatile provider
ok 79 - a directly started compiled host upgrades twice with the same PID, sessions, applied bytes and volatile provider
---
duration_ms: 2795.3936
type: 'test'
...
# Subtest: the running 0.1.0 process upgrades in place and retains its project, session, files and volatile model connection
ok 80 - the running 0.1.0 process upgrades in place and retains its project, session, files and volatile model connection
---
duration_ms: 9319.3218
type: 'test'
...
1..80
# tests 80
# suites 0
# pass 80
# fail 0
# cancelled 0
# skipped 0
# todo 0
# duration_ms 110897.8839
@@ -0,0 +1,20 @@
{
"success": true,
"passed": [
"desktop saves and inspects the full original Task and tool provenance without another model request",
"an outside source change is visible and removes the checkpoint continuation action",
"390px explicit continuation keeps the same Task and one original read; a second resume of the old checkpoint is blocked"
],
"viewportWidths": [
1440,
390
],
"pageErrors": [],
"requestFailures": [],
"screenshots": [
"desktop-original-checkpoint.png",
"mobile-checkpoint-ready.png"
],
"localFixtureModelRequests": 3,
"realModelRequests": 0
}
@@ -0,0 +1,31 @@
{
"success": true,
"version": "1.12.1",
"testedAt": "2026-10-30T15:32:43.531Z",
"directory": "D:\\Ai-project\\REA逆向\\Agent-Workbench\\.qa\\task-plan-ui-848a1079-5b4b-4e62-a5e0-9fc48401d349",
"passed": [
"structured editor seals two real steps and their dependency without running tools",
"original owned read verifies the prerequisite and opens its hashed proof without replay",
"real two-test and build job supplies the current-source proof and completes the sealed plan",
"outside source edits invalidate the current proof, retain the rejection and block a passed delivery report",
"390px immutable revision adds a mobile step, preserves old criteria and keeps evidence dialogs operable",
"a delayed plan-read failure stays with the original task after selection changes",
"controlled tab unmount during a delayed successful save retains the original task result without restoring the closed editor",
"real host restart retains all plan revisions, accepted/rejected proofs and the original operations without replay"
],
"pageErrors": [],
"requestFailures": [],
"screenshots": [
"D:\\Ai-project\\REA逆向\\Agent-Workbench\\.qa\\task-plan-ui-848a1079-5b4b-4e62-a5e0-9fc48401d349\\task-plan-desktop-editor.png",
"D:\\Ai-project\\REA逆向\\Agent-Workbench\\.qa\\task-plan-ui-848a1079-5b4b-4e62-a5e0-9fc48401d349\\task-plan-desktop-verified.png",
"D:\\Ai-project\\REA逆向\\Agent-Workbench\\.qa\\task-plan-ui-848a1079-5b4b-4e62-a5e0-9fc48401d349\\task-plan-mobile-editor.png",
"D:\\Ai-project\\REA逆向\\Agent-Workbench\\.qa\\task-plan-ui-848a1079-5b4b-4e62-a5e0-9fc48401d349\\task-plan-mobile-artifact.png",
"D:\\Ai-project\\REA逆向\\Agent-Workbench\\.qa\\task-plan-ui-848a1079-5b4b-4e62-a5e0-9fc48401d349\\task-plan-mobile-steps.png",
"D:\\Ai-project\\REA逆向\\Agent-Workbench\\.qa\\task-plan-ui-848a1079-5b4b-4e62-a5e0-9fc48401d349\\task-plan-mobile-overview.png"
],
"taskId": "e2113ef4-37ea-496f-be05-ae8a6c0f6489",
"sessionId": "00623235-588b-44c8-a204-5737529075c2",
"planRevisions": 3,
"proofs": 4,
"scope": "Independent real Edge desktop/390px Workbench UI, actual npm checks and owned read fixture. No real model credentials or production service access; semantic project acceptance is separate."
}
@@ -0,0 +1,28 @@
{
"success": true,
"version": "1.12.1",
"at": "2026-10-30T15:35:39.766Z",
"browser": "156.0.4314.50",
"passed": [
"desktop preview seals the committed version and shows existing staged/unstaged/untracked source changes",
"repeated create clicks execute once and preserve the exact source index and owner files",
"opening the child, returning to the source and reloading retain stable project/session identity",
"a record owned by another session navigates to that saved owner instead of accepting its actions",
"explicit recovery preserves outside content and acknowledgement does not report successful creation",
"390×844 preview/create/open/return stays within viewport and preserves source ownership",
"late-failure stays with the original session after project switching",
"late-success stays with the original session after project switching",
"a delayed bootstrap cannot navigate after the task panel has unmounted"
],
"pageErrors": [],
"requestFailures": [],
"screenshots": [
".qa/screenshots/milestone-audit-workspaces-review-desktop.png",
".qa/screenshots/milestone-audit-workspaces-ready-desktop.png",
".qa/screenshots/milestone-audit-workspaces-recovery-desktop.png",
".qa/screenshots/milestone-audit-workspaces-review-mobile.png",
".qa/screenshots/milestone-audit-workspaces-child-mobile.png"
],
"projectRoot": "D:\\Ai-project\\REA逆向\\Agent-Workbench\\.qa\\workspace-ui-ab0158d6-f5bb-4f31-9804-410aa327a027",
"scope": "Independent synthetic Git projects and real browser flows. Desktop/390×844, actual worktree creation, source preservation, explicit external recovery, ownership navigation, delayed success/error/navigation. No production service or real model credential used."
}
+53
View File
@@ -404,3 +404,56 @@ Claude Code 报告已验证工具输入校验、权限决定、队列并发屏
- 正式服务已是1.12.0/PID19084,原Provider和公开hasKey=true状态保持;15项目/26全部会话/2505执行记录/4476事件逐行散列与导出一致。部署后8个桌面/390px页面状态只读复验通过,四类数据库仍不变。物理service-info.appRoot保留最初启动目录,实际runtimeRoot和健康版本才标识已升级运行代码。
- 最终包与测试输入逐份核对,REA-RUNTIME.json仅builtAt不同,其余JSON字段及全部依赖散列严格一致;未广泛忽略清单。209份输入与Git原字节一致,4份旧文件仅CRLF/LF差异且规范化后逐字节相同,未为凑散列改写原文件。
- 6阶段原日志与13份报告散列封存在docs/model-center-release-identity.json;真实模型请求为0,既有用户选择继续有效。桌面壳、手机配对、隔离worker、多协议/多模态和真实工程成功率均为后续独立开发或验收范围。
## 2026-10-30 里程碑审查启动
- 用户要求在推进后续新功能前记录计划、逐里程碑审查和修复漏洞;基线ca9042e/3f6797b、1.12.0独立发行与原证据保留,正式服务不普通重启、不使用真实Key或付费模型。
- 初读发现docs/PRODUCT-PLAN.md开头仍有“1.12源码尚未部署、正式1.11单连接”和“动态MCP未实现”的旧概括,需按最终发行事实更新,同时保留真实模型、WorkBuddy未知格式、OS隔离和桌面/手机等未完成边界。
- 第一份合并补丁因findings.md占位锚点不匹配被拒绝,没有写入;改用实际末段锚点保存。本阶段报告为docs/MILESTONE-AUDIT.md。
- A01已在Windows临时项目实际复现:git-alias/config与credential-alias/config.txt通过内部junction读取受保护夹具,git-alias/new.txt也通过创建预检;直接.git/config被拒绝,说明仅检查词法路径不够。原结果保存.qa/milestone-path-initial-repro.json,未接触用户项目/真实凭据。
- 修复统一safePath:校验已解析真实路径、受限凭据目录后缀和Windows尾点/空格别名;仍允许指向普通项目内文档的链接,未知副作用和原提案核验流程保持。新增真实junction读/列举/提案及Windows路径别名专项验证。
- A02在真实SQLite故障注入中复现:toggle返回磁盘保存失败,但内存enabled=false/revision2,磁盘仍enabled=true/revision1。修复ModelHub保存失败恢复已提交状态;模型选择/运行模式同事务,批量配置导入全部连接、来源与内存凭据统一回滚;原连接设置在成功提交后才更新Key引用。
- npm audit --json --ignore-scripts已完成,153项依赖元数据中0已知漏洞;原输出与散列保存.qa/milestone-dependency-audit。此检查不证明未存在漏洞,也不代替随包REA等依赖身份验证。
- 一次只读查询使用不存在的validate-source-package.mjs,已通过rg确认实际verify-source-package.mjs;没有运行错误的验收器或改写发行证据。
- A03审查热升级发现Provider保留的ModelHub/ConfigImporter对象未更新方法,补丁源码可能无法作用到已初始化模型中心。升级现保留原对象/凭据引用并更新原型,安装失败恢复旧原型;实际在途模型探针/排队会阻止改配置与热升级。新增原对象保留/失败回滚专项。
- A04工作台静态响应原来无frame-ancestors/X-Frame-Options,任意页面可以嵌入带审批能力的本地工作台。补充禁止嵌入、同源脚本/API与no-referrer策略;React仅已有动态布局允许内联style,模型文字仍经安全React文本渲染。实际Edge对照另行验收。
- 首次构建显式调用Node24的npm,但npm脚本仍由原PATH找到Node16,Vite在构建前拒绝;改用本次进程PATH前置Node24,保留失败记录,不更改系统PATH或正式服务。一次不存在Markdown.tsx查询改用实际components.tsx。
- 新接口探针专项首轮19/20,唯一失败在探针成功之后:测试只传mode,违反既有/config完整地址/模型契约导致Invalid URL。前面的并发请求409/0新增派发均已通过;测试改为提交完整配置,不降低原接口验证。模型保存/导入10/10、升级原对象/回滚及两种实际编译宿主升级均通过。
- 接续时复用上一轮源码和原证据,未重复运行初始复现;最后一次版本正则查询因PowerShell引号形成无效正则,改为rg -F明确字符串查询,实际匹配包括verify-change-sets-ui.mjs也需增加1.12.1。一次多份长文读取被限额截断,后续缩小段落/输出。
- 后续顺序已封存docs/NEXT-STEPS.md:先完成审查补丁,再隔离worker、受控引擎及Purge同题对照、Windows桌面、可撤销手机配对、跨协议/多模态、隔离自我迭代。更新总计划的正式1.12与stdio MCP状态,同时保留尚未完成边界。
- A05在隔离子进程中以3个自有环境标记实际复现:PowerShell终端、npm检查及Python检查全部继承宿主未知凭据变量。原证据.qa/milestone-environment-initial-repro.json,未读真实Key。新增projectEnvironment只传系统/运行路径和内核明确字段,三条执行链路统一使用;原Python隔离选项保留。实际shell/npm/Python行为测试验证凭据不继承且UTF-8、测试/构建与结构化证据可用。
- 1.12.1首次完整相关专项80/80、0失败/取消/跳过,原TAP/散列保存.qa/milestone-targeted-673244ae-32ef-46d2-9719-63d73d92598d;修正后的探针API、路径`.`兼容和新增三条真实执行检查均通过。
- 实际旧1.12发行已证明可被外部夹具iframe加载、同进程升级后原ModelHub/Importer、公开配置和会话保持。随后导入故障测试错误地选用无效URL,来源解析按契约预先跳过该项,导致“应拒绝”断言失败;改用预览保留但提交拒绝的超长夹具Key,不改变产品契约。原失败保留.qa/milestone-security-ui-100d0d0e-d705-44a4-a6ae-0d1571541714。
- 修正后真实旧1.12→1.12.1进程、失败批量回滚/预览保留、两类夹具凭据、Edge iframe/外部脚本拒绝、1280/390模型中心和Blob PNG8组通过;所属PID退出、9229关闭,原结果目录9f38a7ec保留。额外配置HTTP磁盘提交故障/非法Key/成功重试检查通过,真实模型请求0。
- 一次task_plan/progress合并补丁使用错误progress首行锚点导致全批拒绝;只读确认两个文件未写入后按实际首行修正,不将补丁失败当作完成。后续冻结前只规范本次修改的代码/测试/脚本LF,原未修改输入不改写。
- 完整UI验收器原gitEnvironment过滤也丢掉独立QA Python路径,现仅额外显式传入该绝对路径,仍不继承模型凭据;完整20类流程进行中。
- 查询归档/便携验收器--help时发现旧脚本未实现帮助分支,实际执行了一次旧1.12独立验收(端口51528、夹具4/4、REA/Edge通过并停止,真实模型0)。只覆盖.qa临时指针,未修改docs历史报告或正式服务。为归档/源码/便携三个入口增加先行帮助和未知参数拒绝;原误运行证据另存,不计为1.12.1验收。
- A01追加静态别名复核实际发现硬链接绕过:AGENTS.md与.secrets/owned.txt共享文件,nlink=2,普通读取和自动项目规则均暴露自有受保护标记。原证据.qa/milestone-hardlink-initial-repro.json,不含真实凭据。realpath无法找出所有硬链接名称,新增常规文件nlink=1保护并在清单跳过多路径文件;普通项目junction仍可用,独立复制文件可正常读取。
- 正在运行的首个独立源码候选20261030150404738保持其原输入,仅作为硬链接补强前基线。新增修复只在工作仓库,不改该解压目录;最终发行需使用追加修复后的新源码/构建/UI与完整回归,不沿用首次候选通过结果。
- 硬链接专项首轮8/9,新增提案测试错误使用proposeFile参数顺序,未达到路径门控即TypeError;已按store/session/args契约修正,原读保护及其余八项通过。未降低保护或改写原失败。
- 首个独立源码候选最终368/369、唯一失败为旧批次Windows别名测试:新的safePath在提案入口已经拒绝尾点,旧测试仍假定先生成提案。更新验证同时要求当前提案入口拒绝,以及旧宿主保留的别名提案仍在批次层拒绝;原失败/TAP与冻结源码目录保留,不登记为最终通过。
- 硬链接读取/提案/清单/规则保护和独立文件副本恢复9/9通过。追加后将重新封存源码及完整UI,首轮150组仅作为补强前基线保存。
- 旧别名提案在批次层也由共用safePath更早拒绝,修正测试首轮仍期待旧“路径别名”文案而失败;按实际保护错误“可读取的项目范围”核对,两层的实际拒绝保留。原失败在执行输出/本记录中,不伪造原TAP。
## 2026-10-30 追加路径复核收尾
- 接续原95492路径/提案/补丁/变更集验证,5文件29/29通过,0失败/取消/跳过;原TAP SHA256为875bc7cebd4893db0468901b4af234b2c684c08a6d52cb5a9d0b53d30ea1176e,副本与原报告保存docs/milestone-audit-path-followup-*。
- 实际硬链接受保护规则可被读取/自动加载的问题已补强,当前多硬链接明确拒绝、独立副本仍可读取;首个独立候选不包含此补强且368/369,不能发布,下一步重新构建和完整验收。
- 一次文档组合补丁因findings占位锚点不存在被拒绝,未写入文档;原证据复制已完成,改用准确锚点及追加记录保存。
- 正式服务当前只读实测为1.12.0/PID23172(2026-10-30T14:33:10Z创建),service-info/appRoot/runtimeRoot均指向原1.12不可变发行;hasKey=true、compatible/gpt-6.1-sol,16项目/20可见会话,无忙任务。历史1.12交付PID19084不是当前进程,本轮未重启服务;原历史数据计数保持为当时记录。升级驱动已绑定新实际preflight,升级前重新快照全部四表/每行及全部会话导出,保留当前Provider与新增数据,未读取Key/发送模型请求。
- 7b1f529新构建完整UI20类150组全部通过、0页面/请求异常,正式4318不参与夹具;当前源码快照218份输入/6份构建已保存.qa/milestone-final-code-baseline.json。首轮候选证据已另存,接下来生成新的不可变包并独立完整回归,不重复开发树全量。
+20
View File
@@ -1,5 +1,15 @@
# 统一执行进度
## 2026-10-30 里程碑审查与1.12.1候选
- 后续顺序、前置和真实完成门槛已保存docs/NEXT-STEPS.md;docs/MILESTONE-AUDIT.md逐版本对照1.0–1.12,并明确M0–M6哪些只是部分完成。更新PRODUCT-PLAN与VALIDATION的正式1.12/stdio MCP和1.7–1.9索引。
- A01路径、A02配置事务、A03保留模型原型/活动门控、A04页面策略和A05宿主环境传递已修复;80/80相关专项与额外实际配置HTTP故障检查通过。
- 原不可变1.12进程的实际升级与Edge8组通过:原PID、初始化Hub/Importer、预览和夹具凭据保留;失败批量不部分导入;第三方iframe/脚本拒绝,1280/390模型中心与Blob PNG仍正常。原证据.qa/milestone-security-ui-9f38a7ec-8649-4260-9c43-375425e6c8dd,两张截图已目视核对。
- 首轮Edge测试选择无效URL作为提交故障,但预览按契约跳过,原失败目录100d0d0e保留;修正为超长自有凭据后通过。一次含错误progress首行锚点的合并补丁被拒绝,确认所有目标文件未写入,再按实际首行应用。
- 准备完整UI与独立源码候选验收。正式4318暂未升级、不普通重启、不读取真实Key、真实模型0请求;后续隔离worker等新能力尚未开始。
- 完整20类150组UI全部通过,0页面/请求异常;源代码修复已分步提交c1e9575/f4bd0ad。24项能力基线已更新至实际交付1.12,原总计划与历史执行记录分开;后续冻结独立源码回归,期间不修改实现输入。
## 2026-10-30 MCP管理与运行记录接续
- 接续 `feat/model-center-1.12` 工作树;Skills/检查点提交已存在,MCP 改动仍未提交且完整保留。
@@ -458,3 +468,13 @@
- 部署后1280px/390px各检查执行计划、模型中心、Skills和MCP,8个实际页面状态通过,0页面异常/请求失败/HTTP错误/写入请求;四类数据库散列仍一致。6阶段日志和13份报告已封存docs/model-center-release-identity.json,调试端口9229关闭,本轮真实模型请求0。
- 升级前首次REA清单时间差使流程在新handler安装前停止;部署后Python环境缺Playwright使检查在浏览器启动前退出,随后使用锁定Node Playwright完成8组只读复验。两次均保留原记录,不修改产品代码、不重放工具或调用模型。
- 本轮收尾仅修正文档中的旧候选/未发布状态,补充最终验收和后续开发顺序。代码已由3f6797b封存;最终报告和规划另作本地文档提交,不推送。下一独立阶段为隔离worker及受控外部引擎,再推进桌面、真实手机配对、多协议/多模态与迭代评测;真实模型验收继续按用户选择暂缓。
## 2026-10-30 追加路径复核收尾
- 接续原95492路径/提案/补丁/变更集验证,5文件29/29通过,0失败/取消/跳过;原TAP SHA256为875bc7cebd4893db0468901b4af234b2c684c08a6d52cb5a9d0b53d30ea1176e,副本与原报告保存docs/milestone-audit-path-followup-*。
- 实际硬链接受保护规则可被读取/自动加载的问题已补强,当前多硬链接明确拒绝、独立副本仍可读取;首个独立候选不包含此补强且368/369,不能发布,下一步重新构建和完整验收。
- 一次文档组合补丁因findings占位锚点不存在被拒绝,未写入文档;原证据复制已完成,改用准确锚点及追加记录保存。
- 正式服务当前只读实测为1.12.0/PID23172(2026-10-30T14:33:10Z创建),service-info/appRoot/runtimeRoot均指向原1.12不可变发行;hasKey=true、compatible/gpt-6.1-sol,16项目/20可见会话,无忙任务。历史1.12交付PID19084不是当前进程,本轮未重启服务;原历史数据计数保持为当时记录。升级驱动已绑定新实际preflight,升级前重新快照全部四表/每行及全部会话导出,保留当前Provider与新增数据,未读取Key/发送模型请求。
- 7b1f529新构建完整UI20类150组全部通过、0页面/请求异常,正式4318不参与夹具;当前源码快照218份输入/6份构建已保存.qa/milestone-final-code-baseline.json。首轮候选证据已另存,接下来生成新的不可变包并独立完整回归,不重复开发树全量。
+12
View File
@@ -1,7 +1,19 @@
# Agent Workbench:核心代理能力研究与产品计划
## 2026-10-30 当前任务:里程碑回顾、漏洞修复与后续计划封存
用户要求先回顾审查每个里程碑、修复漏洞并核对计划实现,再推进后续新能力。本阶段暂停新功能切片,不暂停已部署1.12.0服务;真实模型验收继续暂缓。
1. [complete] 后续顺序/前置/验收门槛已写docs/NEXT-STEPS.md;1.0–1.12全部节点及M0–M6计划/实现/原证据/遗留项已对照,修正总计划/验证索引过时概括。
2. [complete] 核对HTTP、路径/Git、执行/取消/恢复、上下文/计划、模型/导入、Skills/MCP/REA/引擎/浏览器;A01/A02/A05真实隔离复现已封存,依赖0已知漏洞另记。
3. [complete] A01–A05修复及追加硬链接补强完成;80项相关专项、额外配置故障HTTP检查、实际旧1.12→补丁升级/Edge策略8组及路径/提案/变更集29项复验通过;初始复现与失败保留。两张新界面/拒绝嵌入截图已目视核对。
4. [in_progress] 首轮1.12.1独立源码368/369,唯一失败为Windows尾点已在提案层提前拒绝而旧测试仍期待生成提案;更新当前/旧提案两层断言,29项复验通过。7b1f529与新构建的20类150组UI全部通过,218份输入/6份构建已快照;进入新不可变候选、独立源码冻结回归和保留当前正式Provider/全部历史验收;首失败候选不能部署。
5. [pending] 交付审查报告、修复与复验结果、未完成能力和后续起点。真实模型请求为0,后续隔离worker等只在本阶段收尾后推进。
## 2026-10-30 当前接续:1.12 完整发行与本地交付
以下按原过程保留历史执行记录;历史中的pending/in_progress不作为新的执行队列。最新阶段状态以本文顶部为准,尚未完成的能力与后续顺序统一见docs/NEXT-STEPS.md。
1. [complete] 模型中心/池、来源核对记忆、检查点、Skills已本地实现并分段提交;MCP实际stdio、未知结果门控、人工核对与无报告恢复已补强。
2. [complete] MCP补强后20项专项、6组管理与2组真实中断界面通过;353项旧全量仅作为补强前基线,不替代最终源码结果。
3. [complete] 统一1.12版本/文档/验收脚本;20类150组界面通过,标题栏5组复验及最终MCP6+2组复验通过。修复并实际复验长MCP参数审批上限,20项stdio/注册专项全通过。