<!-- 使用 release/v0.21.4-nightly.20260804.d6f55a1c9 分支中 .github/release.yml 的配置生成发布说明 -->
## 更新内容
* feat(desktop): 引导 Electron 用户迁移至 Tauri 更新 by @yiliang114 in https://github.com/QwenLM/qwen-code/pull/8392
* fix(web-shell): 防止表格对话框关闭时滚动位置跳动 by @carffuca in https://github.com/QwenLM/qwen-code/pull/8407
* ci: 移除损坏的旧版定时 PR 分类工作流 by @wenshao in https://github.com/QwenLM/qwen-code/pull/8434
* feat(telemetry): 追踪工具执行结果 by @doudouOUC in https://github.com/QwenLM/qwen-code/pull/8180
* fix(core): 明确 write_file 的预读取指导 by @doudouOUC in https://github.com/QwenLM/qwen-code/pull/8428
* fix(desktop): 从 LocalAppData 读取 Windows 冒烟测试日志 by @yiliang114 in https://github.com/QwenLM/qwen-code/pull/8381
* fix(core): 避免重放不安全的 MCP 工具调用 by @doudouOUC in https://github.com/QwenLM/qwen-code/pull/8387
* fix(core): 在证据目录耗尽后停止目标重试 by @qqqys in https://github.com/QwenLM/qwen-code/pull/8430
* fix(review): 将已交付但发生漂移的发布排除在修复循环之外 by @wenshao in https://github.com/QwenLM/qwen-code/pull/8466
* feat(review): 添加结构化的 Web Shell 审查结果 by @wenshao in https://github.com/QwenLM/qwen-code/pull/8402
* ci: 减少 SDK Java 运行器的排队 by @yiliang114 in https://github.com/QwenLM/qwen-code/pull/8441
* fix(core): 使 MCP 重连超时测试与安全重放策略保持一致 by @wenshao in https://github.com/QwenLM/qwen-code/pull/8478
* feat(core): 支持 Qwen 3.8 推理强度 by @DragonnZhang in https://github.com/QwenLM/qwen-code/pull/8472
* fix(ci): 使审查工作流测试与外部化超时变量保持一致 by @wenshao in https://github.com/QwenLM/qwen-code/pull/8486
**完整更新日志**: https://github.com/QwenLM/qwen-code/compare/v0.21.4...v0.21.4-nightly.20260804.d6f55a1c9
<!-- Release notes generated using configuration in .github/release.yml at release/v0.21.4-nightly.20260804.d6f55a1c9 -->
## What's Changed
* feat(desktop): bridge Electron users to Tauri updates by @yiliang114 in https://github.com/QwenLM/qwen-code/pull/8392
* fix(web-shell): prevent table dialog close scroll jump by @carffuca in https://github.com/QwenLM/qwen-code/pull/8407
* ci: remove broken legacy scheduled PR triage workflow by @wenshao in https://github.com/QwenLM/qwen-code/pull/8434
* feat(telemetry): Track tool execution outcomes by @doudouOUC in https://github.com/QwenLM/qwen-code/pull/8180
* fix(core): Clarify write_file prior-read guidance by @doudouOUC in https://github.com/QwenLM/qwen-code/pull/8428
* fix(desktop): read Windows smoke log from LocalAppData by @yiliang114 in https://github.com/QwenLM/qwen-code/pull/8381
* fix(core): Avoid replaying unsafe MCP tool calls by @doudouOUC in https://github.com/QwenLM/qwen-code/pull/8387
* fix(core): stop Goal retries after evidence catalog exhaustion by @qqqys in https://github.com/QwenLM/qwen-code/pull/8430
* fix(review): keep drifted-but-delivered launches out of the repair loop by @wenshao in https://github.com/QwenLM/qwen-code/pull/8466
* feat(review): Add structured Web Shell review results by @wenshao in https://github.com/QwenLM/qwen-code/pull/8402
* ci: reduce SDK Java runner queueing by @yiliang114 in https://github.com/QwenLM/qwen-code/pull/8441
* fix(core): align MCP reconnect timeout test with safe replay policy by @wenshao in https://github.com/QwenLM/qwen-code/pull/8478
* feat(core): support Qwen 3.8 reasoning effort by @DragonnZhang in https://github.com/QwenLM/qwen-code/pull/8472
* fix(ci): align review workflow tests with externalized timeout variables by @wenshao in https://github.com/QwenLM/qwen-code/pull/8486
**Full Changelog**: https://github.com/QwenLM/qwen-code/compare/v0.21.4...v0.21.4-nightly.20260804.d6f55a1c9