qdaxb · Jul 18, 2026 · Jul 18, 2026
diff --git a/docs/en/user-guide/coding/managing-code-tasks.md b/docs/en/user-guide/coding/managing-code-tasks.md
@@ -25,19 +25,20 @@ This guide explains how to create, execute, and manage code-type tasks in Wegent
 A code task is a task type in Wegent specifically designed for software development. Unlike regular chat tasks, code tasks connect to Git repositories, allowing AI agents to make code changes directly in the repository.
 
 **Core Concept**:
+
 ```
 Code Task = User Prompt + Code Agent + Git Repository + Branch
 ```
 
 ### Code Tasks vs Chat Tasks
 
-| Feature | Code Task | Chat Task |
-|---------|-----------|-----------|
-| **Git Repository** | Required | Optional |
-| **Code Execution** | Runs in Docker container | No code execution |
-| **Workbench** | Shows file changes, commit history | Not displayed |
-| **Branch Management** | Auto-creates feature branches | None |
-| **Use Cases** | Development, refactoring, bug fixes | Q&A, analysis, documentation |
+| Feature               | Code Task                           | Chat Task                    |
+| --------------------- | ----------------------------------- | ---------------------------- |
+| **Git Repository**    | Required                            | Optional                     |
+| **Code Execution**    | Runs in Docker container            | No code execution            |
+| **Workbench**         | Shows file changes, commit history  | Not displayed                |
+| **Branch Management** | Auto-creates feature branches       | None                         |
+| **Use Cases**         | Development, refactoring, bug fixes | Q&A, analysis, documentation |
 
 ---
 
@@ -146,13 +147,13 @@ graph LR
 
 ### Task States
 
-| Status | Description | Actions |
-|--------|-------------|---------|
-| **PENDING** | Waiting to execute | Can cancel |
-| **RUNNING** | Currently executing | Can stop |
-| **COMPLETED** | Execution finished | Can view results, create PR |
-| **FAILED** | Execution failed | Can retry |
-| **CANCELLED** | Was cancelled | Can recreate |
+| Status        | Description         | Actions                     |
+| ------------- | ------------------- | --------------------------- |
+| **PENDING**   | Waiting to execute  | Can cancel                  |
+| **RUNNING**   | Currently executing | Can stop                    |
+| **COMPLETED** | Execution finished  | Can view results, create PR |
+| **FAILED**    | Execution failed    | Can retry                   |
+| **CANCELLED** | Was cancelled       | Can recreate                |
 
 ### Stopping a Task
 
@@ -187,6 +188,7 @@ After task completion, you can continue chatting with the agent:
 View detailed execution information in the Workbench:
 
 - **Execution Timeline**: See tools used by AI and execution order
+- **Tool Duration**: Shows seconds below one minute, then minutes and seconds from one minute onward
 - **Commit History**: View all code commits
 - **File Changes**: See specific modifications for each file
 
@@ -235,49 +237,57 @@ For API details, see [Runtime Cleanup](../../developer-guide/runtime-cleanup.md)
 ### Q1: Task stuck in PENDING status?
 
 **Possible causes**:
+
 1. No available execution containers
 2. Repository access permission issues
 3. Git token expired
 
 **Solutions**:
+
 - Check if system resources are sufficient
 - Verify Git token is valid
 - Check repository access permissions
 
 ### Q2: Code commit failed?
 
 **Possible causes**:
+
 1. Branch protection rules
 2. Insufficient permissions
 3. Network issues
 
 **Solutions**:
+
 - Check target branch protection rules
 - Confirm Git token has write permissions
 - Retry the task
 
 ### Q3: AI modified the wrong files?
 
 **Solutions**:
+
 1. Explicitly specify file paths to modify in the task
 2. Provide more detailed context information
 3. Use knowledge base to provide project structure documentation
 
 ### Q4: How to make AI follow project coding standards?
 
 **Solutions**:
+
 1. Add `.cursorrules` or `.windsurfrules` file in repository root
 2. Explicitly state coding standards in task description
 3. Use knowledge base to provide coding standards documentation
 
 ### Q5: Task execution taking too long?
 
 **Possible causes**:
+
 1. Task scope too large
 2. Need to install many dependencies
 3. Network latency
 
 **Solutions**:
+
 - Split large tasks into smaller ones
 - Use pre-configured base images
 - Check network connection

diff --git a/docs/zh/user-guide/coding/managing-code-tasks.md b/docs/zh/user-guide/coding/managing-code-tasks.md
@@ -25,19 +25,20 @@ sidebar_position: 2
 代码任务是 Wegent 中专门用于软件开发的任务类型。与普通聊天任务不同，代码任务会连接到 Git 仓库，AI 智能体可以直接在仓库中进行代码修改。
 
 **核心概念**：
+
 ```
 代码任务 = 用户提示词 + 代码智能体 + Git 仓库 + 分支
 ```
 
 ### 代码任务 vs 聊天任务
 
-| 特性 | 代码任务 | 聊天任务 |
-|------|----------|----------|
-| **Git 仓库** | 必需 | 可选 |
-| **代码执行** | 在 Docker 容器中执行 | 无代码执行 |
-| **工作台** | 显示文件变更、提交历史 | 不显示 |
-| **分支管理** | 自动创建功能分支 | 无 |
-| **适用场景** | 代码开发、重构、修复 | 问答、分析、文档 |
+| 特性         | 代码任务               | 聊天任务         |
+| ------------ | ---------------------- | ---------------- |
+| **Git 仓库** | 必需                   | 可选             |
+| **代码执行** | 在 Docker 容器中执行   | 无代码执行       |
+| **工作台**   | 显示文件变更、提交历史 | 不显示           |
+| **分支管理** | 自动创建功能分支       | 无               |
+| **适用场景** | 代码开发、重构、修复   | 问答、分析、文档 |
 
 ---
 
@@ -146,13 +147,13 @@ graph LR
 
 ### 任务状态
 
-| 状态 | 描述 | 操作 |
-|------|------|------|
-| **PENDING** | 等待执行 | 可取消 |
-| **RUNNING** | 正在执行 | 可停止 |
+| 状态          | 描述     | 操作                |
+| ------------- | -------- | ------------------- |
+| **PENDING**   | 等待执行 | 可取消              |
+| **RUNNING**   | 正在执行 | 可停止              |
 | **COMPLETED** | 执行完成 | 可查看结果、创建 PR |
-| **FAILED** | 执行失败 | 可重试 |
-| **CANCELLED** | 已取消 | 可重新创建 |
+| **FAILED**    | 执行失败 | 可重试              |
+| **CANCELLED** | 已取消   | 可重新创建          |
 
 ### 停止任务
 
@@ -187,6 +188,7 @@ graph LR
 在工作台中查看详细的执行信息：
 
 - **执行时间线**：查看 AI 使用的工具和执行顺序
+- **工具耗时**：不足 1 分钟时按秒显示，达到 1 分钟后按“分 + 秒”显示
 - **提交历史**：查看所有代码提交
 - **文件变更**：查看每个文件的具体修改
 
@@ -235,49 +237,57 @@ graph LR
 ### Q1：任务一直处于 PENDING 状态？
 
 **可能原因**：
+
 1. 没有可用的执行容器
 2. 仓库访问权限问题
 3. Git 令牌过期
 
 **解决方案**：
+
 - 检查系统资源是否充足
 - 验证 Git 令牌是否有效
 - 检查仓库访问权限
 
 ### Q2：代码提交失败？
 
 **可能原因**：
+
 1. 分支保护规则
 2. 权限不足
 3. 网络问题
 
 **解决方案**：
+
 - 检查目标分支的保护规则
 - 确认 Git 令牌有写入权限
 - 重试任务
 
 ### Q3：AI 修改了错误的文件？
 
 **解决方案**：
+
 1. 在任务中明确指定要修改的文件路径
 2. 提供更详细的上下文信息
 3. 使用知识库提供项目结构说明
 
 ### Q4：如何让 AI 遵循项目的编码规范？
 
 **解决方案**：
+
 1. 在仓库根目录添加 `.cursorrules` 或 `.windsurfrules` 文件
 2. 在任务描述中明确说明编码规范
 3. 使用知识库提供编码规范文档
 
 ### Q5：任务执行时间过长？
 
 **可能原因**：
+
 1. 任务范围过大
 2. 需要安装大量依赖
 3. 网络延迟
 
 **解决方案**：
+
 - 将大任务拆分成小任务
 - 使用预配置的基础镜像
 - 检查网络连接

diff --git a/wework/src/components/chat/blocks/ToolBlocksDisplay.test.tsx b/wework/src/components/chat/blocks/ToolBlocksDisplay.test.tsx
@@ -761,7 +761,7 @@ describe('ToolBlocksDisplay', () => {
     expect(screen.getByRole('button', { name: /已处理 6 分 8 秒/ })).toBeInTheDocument()
   })
 
-  test('formats live duration with natural Chinese units', () => {
+  test('formats live duration with minutes and seconds after one minute', () => {
     vi.useFakeTimers()
     vi.setSystemTime(new Date('2026-06-05T00:00:00.000Z'))
 
@@ -777,7 +777,26 @@ describe('ToolBlocksDisplay', () => {
       vi.advanceTimersByTime(62000)
     })
 
-    expect(screen.getByText('62 秒')).toBeInTheDocument()
+    expect(screen.getByText('1 分 2 秒')).toBeInTheDocument()
+  })
+
+  test('formats exactly one minute with zero remaining seconds', () => {
+    vi.useFakeTimers()
+    vi.setSystemTime(new Date('2026-06-05T00:00:00.000Z'))
+
+    const runningBlock: ProcessingBlock = {
+      ...completedCommandBlock,
+      status: 'streaming',
+      createdAt: Date.now(),
+    }
+
+    render(<ToolBlocksDisplay blocks={[runningBlock]} isStreaming={true} />)
+
+    act(() => {
+      vi.advanceTimersByTime(60000)
+    })
+
+    expect(screen.getByText('1 分 0 秒')).toBeInTheDocument()
   })
 
   test('keeps ticking while streaming even when all tool blocks are done', () => {

diff --git a/wework/src/components/chat/blocks/processingDuration.ts b/wework/src/components/chat/blocks/processingDuration.ts
@@ -20,7 +20,11 @@ export function getWholeSecondsDurationText(
   isRunning: boolean
 ): string {
   const durationMs = getProcessingDurationMs(blocks, turnStartedAt, now, completedAt, isRunning)
-  return `${Math.floor(durationMs / 1000)} 秒`
+  const seconds = Math.floor(durationMs / 1000)
+  if (seconds < 60) return `${seconds} 秒`
+
+  const minutes = Math.floor(seconds / 60)
+  return `${minutes} 分 ${seconds % 60} 秒`
 }
 
 function getProcessingDurationMs(