<!-- 使用 release/v0.22.0-nightly.20260824.3a1f86d805 分支中 .github/release.yml 的配置生成 -->
## 更新内容
* 修复(web-shell)：从概览面板打开时传递会话工作区的当前工作目录 (cwd) by @wenshao in https://github.com/QwenLM/qwen-code/pull/9730
* 修复(web-shell)：在过时的中途拒绝后回退 by @ytahdn in https://github.com/QwenLM/qwen-code/pull/9732
* 修复(review)：阻止回退评论在已废弃的运行中触发 by @wenshao in https://github.com/QwenLM/qwen-code/pull/9716
* 功能(review)：添加时间可达性与事件回放视图 by @wenshao in https://github.com/QwenLM/qwen-code/pull/9708
* 修复(web-shell)：稳定固定区域的顺序并提供即时固定反馈 by @yiliang114 in https://github.com/QwenLM/qwen-code/pull/9560
* 功能(review)：要求每个修复提供其测试，并对不收敛的情况做出裁决 by @wenshao in https://github.com/QwenLM/qwen-code/pull/9596
* 维护(ci)：将自动修复说明迁移至设计记录并遏制增长 by @wenshao in https://github.com/QwenLM/qwen-code/pull/9677
* 功能(cli)：将每个守护进程会话恢复至其上次选择的模型 by @doudouOUC in https://github.com/QwenLM/qwen-code/pull/9687
* 功能(review)：在发布前根据捕获的差异验证 Aone 行内锚点 by @wenshao in https://github.com/QwenLM/qwen-code/pull/9634
* 修复(ci)：在 workflow 尺寸基线中记录 qwen-autofix.yml 的交付尺寸 by @wenshao in https://github.com/QwenLM/qwen-code/pull/9747
* 修复(ci)：根据 review 载荷提供的字段限制 fork 信号 by @qqqys in https://github.com/QwenLM/qwen-code/pull/9469
* 功能(goal)：核算 Goal 消耗的 token by @qqqys in https://github.com/QwenLM/qwen-code/pull/9583
* 修复(core)：拒绝回显提示词自身示例的会话标题 by @yiliang114 in https://github.com/QwenLM/qwen-code/pull/9709
* 修复(core)：解析 workflow 元字面量而非直接求值 by @qqqys in https://github.com/QwenLM/qwen-code/pull/9325
* 修复(sdk)：支持相对路径的制品下载 URL by @ytahdn in https://github.com/QwenLM/qwen-code/pull/9734
* 修复(autofix)：为修复轮次提供可完成的预算 by @qqqys in https://github.com/QwenLM/qwen-code/pull/9691
* 功能(cli)：通过设置键启用动态 workflow by @qqqys in https://github.com/QwenLM/qwen-code/pull/9098
* 修复(core)：将 workflow 停滞看门狗的范围扩展至传输重试链之外 by @qqqys in https://github.com/QwenLM/qwen-code/pull/9397
* 测试(core)：验证子代理注册表不包含任何 MCP 指令 by @wenshao in https://github.com/QwenLM/qwen-code/pull/9720
* 修复(core)：根据模型声明的内容限制技能公告 by @wenshao in https://github.com/QwenLM/qwen-code/pull/9718
* 功能(review)：为 Aone Code 目标提供 pr-context 支持 by @wenshao in https://github.com/QwenLM/qwen-code/pull/9621
* 修复(review)：将修复引发的重新报告计为首次工作 by @wenshao in https://github.com/QwenLM/qwen-code/pull/9744
* 文档(review)：当两个轴都称为“rank 3”时为其命名 by @wenshao in https://github.com/QwenLM/qwen-code/pull/9759
* 修复(web-shell)：预览按文档分类的制品 by @ytahdn in https://github.com/QwenLM/qwen-code/pull/9760
* 功能(web-shell)：在切换后增量刷新 composer 技能 by @samuelhsin in https://github.com/QwenLM/qwen-code/pull/9131
* 功能(review)：为 Aone Code 目标提供 comment-status 和 presubmit 支持 by @wenshao in https://github.com/QwenLM/qwen-code/pull/9627
* 性能(cli)：将 VP 滚动渲染提升至 60 FPS by @DragonnZhang in https://github.com/QwenLM/qwen-code/pull/9681
* 功能(models)：支持双角色图像生成模型 by @callmeYe in https://github.com/QwenLM/qwen-code/pull/9650
* 修复(daemon)：加载后保持恢复的 ask_user_question 有效 by @doudouOUC in https://github.com/QwenLM/qwen-code/pull/9763
* 功能(review)：添加持续关键收敛建议（带残余风险合并） by @wenshao in https://github.com/QwenLM/qwen-code/pull/9526
* 修复(core)：支持使用旧版 Git 的公共 GitHub 扩展 by @yiliang114 in https://github.com/QwenLM/qwen-code/pull/9690
* 功能(cua-driver)：添加带版本控制的 Computer Use SDK 及发布流水线 by @LaZzyMan in https://github.com/QwenLM/qwen-code/pull/9587
* 重构(node-repl)!: 将持久化 Node REPL 作为独立 MCP 服务器交付 by @LaZzyMan in https://github.com/QwenLM/qwen-code/pull/9499
* 修复(review)：在放弃 worktree 清理前修复权限 by @wenshao in https://github.com/QwenLM/qwen-code/pull/9748
* 修复(core)：规范化会话标题回显比较的两侧 by @yiliang114 in https://github.com/QwenLM/qwen-code/pull/9809
* 重构(cli)：强制执行 utils 叶层依赖方向 (#9146) by @yiliang114 in https://github.com/QwenLM/qwen-code/pull/9737
* 修复(web-shell)：限制 React dev performance.measure 的累积以防止渲染器内存溢出 by @wenshao in https://github.com/QwenLM/qwen-code/pull/9770
* 修复(dingtalk)：解析转发聊天记录 by @qqqys in https://github.com/QwenLM/qwen-code/pull/9339
* 修复(web-shell)：添加 remark-cjk-friendly 以解决 CJK 标点符号旁加粗的问题… by @cactuser-Lu in https://github.com/QwenLM/qwen-code/pull/9457
* 功能(mcp)：添加 MCP 2026 核心及 WebShell Apps 主机 by @samuelhsin in https://github.com/QwenLM/qwen-code/pull/8992
* 修复(config)：在设置模式中允许 prompt hooks by @zjunothing in https://github.com/QwenLM/qwen-code/pull/8779
* 修复(web-shell)：在会话创建前显示推理强度 by @callmeYe in https://github.com/QwenLM/qwen-code/pull/9599
* 修复(auth)：使 Vertex AI 能够通过应用默认凭据进行认证 by @axiom-of-choice in https://github.com/QwenLM/qwen-code/pull/9017
* 修复(ci)：在 workflow 尺寸基线中记录 cd-cua-driver.yml 的交付尺寸 by @wenshao in https://github.com/QwenLM/qwen-code/pull/9822
* 功能(review)：为验证器提供不可反驳列表及可构建的拒绝阈值 by @wenshao in https://github.com/QwenLM/qwen-code/pull/9799
## 新贡献者
* @cactuser-Lu 在 https://github.com/QwenLM/qwen-code/pull/9457 中首次贡献
* @axiom-of-choice 在 https://github.com/QwenLM/qwen-code/pull/9017 中首次贡献
**完整更新日志**: https://github.com/QwenLM/qwen-code/compare/v0.22.0...v0.22.0-nightly.20260824.3a1f86d805
<!-- Release notes generated using configuration in .github/release.yml at release/v0.22.0-nightly.20260824.3a1f86d805 -->
## What's Changed
* fix(web-shell): pass session workspace cwd when opening from overview panel by @wenshao in https://github.com/QwenLM/qwen-code/pull/9730
* fix(web-shell): fall back after stale mid-turn rejection by @ytahdn in https://github.com/QwenLM/qwen-code/pull/9732
* fix(review): stop the fallback comment from firing on superseded runs by @wenshao in https://github.com/QwenLM/qwen-code/pull/9716
* feat(review): add temporal-reachability and incident-replay lenses by @wenshao in https://github.com/QwenLM/qwen-code/pull/9708
* fix(web-shell): stable pinned-section order and instant pin feedback by @yiliang114 in https://github.com/QwenLM/qwen-code/pull/9560
* feat(review): ask each fix for its test, and rule on non-convergence by @wenshao in https://github.com/QwenLM/qwen-code/pull/9596
* chore(ci): migrate autofix prose to the design record and ratchet growth by @wenshao in https://github.com/QwenLM/qwen-code/pull/9677
* feat(cli): restore each daemon session onto its last selected model by @doudouOUC in https://github.com/QwenLM/qwen-code/pull/9687
* feat(review): validate Aone inline anchors against the captured diff before posting by @wenshao in https://github.com/QwenLM/qwen-code/pull/9634
* fix(ci): record qwen-autofix.yml's shipped size in the workflow size baseline by @wenshao in https://github.com/QwenLM/qwen-code/pull/9747
* fix(ci): gate the fork signal on fields the review payload delivers by @qqqys in https://github.com/QwenLM/qwen-code/pull/9469
* feat(goal): account the tokens a Goal spends by @qqqys in https://github.com/QwenLM/qwen-code/pull/9583
* fix(core): reject session titles that echo the prompt's own examples by @yiliang114 in https://github.com/QwenLM/qwen-code/pull/9709
* fix(core): parse the workflow meta literal instead of evaluating it by @qqqys in https://github.com/QwenLM/qwen-code/pull/9325
* fix(sdk): support relative artifact download URLs by @ytahdn in https://github.com/QwenLM/qwen-code/pull/9734
* fix(autofix): give the repair pass a budget it can finish in by @qqqys in https://github.com/QwenLM/qwen-code/pull/9691
* feat(cli): enable dynamic workflows from a settings key by @qqqys in https://github.com/QwenLM/qwen-code/pull/9098
* fix(core): widen the workflow stall watchdog past the transport retry ladder by @qqqys in https://github.com/QwenLM/qwen-code/pull/9397
* test(core): pin that subagent registries carry no MCP instructions by @wenshao in https://github.com/QwenLM/qwen-code/pull/9720
* fix(core): gate skill announcements on what the model was declared by @wenshao in https://github.com/QwenLM/qwen-code/pull/9718
* feat(review): back pr-context on Aone Code targets by @wenshao in https://github.com/QwenLM/qwen-code/pull/9621
* fix(review): count a fix-induced re-report as first-time work by @wenshao in https://github.com/QwenLM/qwen-code/pull/9744
* docs(review): name the axis when two of them both call it "rank 3" by @wenshao in https://github.com/QwenLM/qwen-code/pull/9759
* fix(web-shell): preview document-classified artifacts by @ytahdn in https://github.com/QwenLM/qwen-code/pull/9760
* feat(web-shell): refresh composer skills incrementally after toggles by @samuelhsin in https://github.com/QwenLM/qwen-code/pull/9131
* feat(review): back comment-status and presubmit for Aone Code targets by @wenshao in https://github.com/QwenLM/qwen-code/pull/9627
* perf(cli): raise VP scroll rendering to 60 FPS by @DragonnZhang in https://github.com/QwenLM/qwen-code/pull/9681
* feat(models): support dual-role image generation models by @callmeYe in https://github.com/QwenLM/qwen-code/pull/9650
* fix(daemon): keep restored ask_user_question valid after load by @doudouOUC in https://github.com/QwenLM/qwen-code/pull/9763
* feat(review): add the persistently-critical convergence advisory (land-with-residual-risk) by @wenshao in https://github.com/QwenLM/qwen-code/pull/9526
* fix(core): support public GitHub extensions with older Git by @yiliang114 in https://github.com/QwenLM/qwen-code/pull/9690
* feat(cua-driver): add versioned Computer Use SDK and release pipeline by @LaZzyMan in https://github.com/QwenLM/qwen-code/pull/9587
* refactor(node-repl)!: deliver the persistent Node REPL as a standalone MCP server by @LaZzyMan in https://github.com/QwenLM/qwen-code/pull/9499
* fix(review): repair permissions before giving up on worktree cleanup by @wenshao in https://github.com/QwenLM/qwen-code/pull/9748
* fix(core): Normalize both sides of the session title echo comparison by @yiliang114 in https://github.com/QwenLM/qwen-code/pull/9809
* refactor(cli): enforce utils leaf-layer dependency direction (#9146) by @yiliang114 in https://github.com/QwenLM/qwen-code/pull/9737
* fix(web-shell): cap React dev performance.measure accumulation to stop renderer OOM by @wenshao in https://github.com/QwenLM/qwen-code/pull/9770
* fix(dingtalk): parse forwarded chat records by @qqqys in https://github.com/QwenLM/qwen-code/pull/9339
* fix(web-shell): add remark-cjk-friendly so bold adjacent to CJK punct… by @cactuser-Lu in https://github.com/QwenLM/qwen-code/pull/9457
* feat(mcp): add MCP 2026 core and WebShell Apps host by @samuelhsin in https://github.com/QwenLM/qwen-code/pull/8992
* fix(config): allow prompt hooks in settings schema by @zjunothing in https://github.com/QwenLM/qwen-code/pull/8779
* fix(web-shell): show reasoning effort before session creation by @callmeYe in https://github.com/QwenLM/qwen-code/pull/9599
* fix(auth): let Vertex AI authenticate with Application Default Credentials by @axiom-of-choice in https://github.com/QwenLM/qwen-code/pull/9017
* fix(ci): record cd-cua-driver.yml's shipped size in the workflow size baseline by @wenshao in https://github.com/QwenLM/qwen-code/pull/9822
* feat(review): give verifiers a do-not-refute list and a constructible rejection bar by @wenshao in https://github.com/QwenLM/qwen-code/pull/9799
## New Contributors
* @cactuser-Lu made their first contribution in https://github.com/QwenLM/qwen-code/pull/9457
* @axiom-of-choice made their first contribution in https://github.com/QwenLM/qwen-code/pull/9017
**Full Changelog**: https://github.com/QwenLM/qwen-code/compare/v0.22.0...v0.22.0-nightly.20260824.3a1f86d805