Claude Team MCP

Instructions

Implementation Reference

• src/server.ts:871-920 (handler)

  Main execution handler for the 'run_workflow' tool. Fetches the workflow definition, plans experts (for logging), runs orchestrator.execute, saves history, and returns results with progress.
  case 'run_workflow': { const { workflow: workflowId, task, context } = args as { workflow: string; task: string; context?: string }; const workflow = workflowManager.getWorkflow(workflowId); if (!workflow) { throw new Error(`工作流不存在: ${workflowId}`); } // 转换为 Tech Lead 格式 const { experts } = workflowManager.toTaskAnalysis(workflow, task); const startTime = Date.now(); const progressLogs: string[] = []; orchestrator.setProgressCallback((message, progress) => { const timestamp = new Date().toLocaleTimeString(); const progressStr = progress ? ` (${progress}%)` : ''; progressLogs.push(`[${timestamp}]${progressStr} ${message}`); }); // 使用工作流执行任务 progressLogs.push(`📋 使用工作流: ${workflow.name}`); progressLogs.push(`👥 创建 ${experts.length} 位专家: ${experts.map((e: { name: string }) => e.name).join(', ')}`); const result = await orchestrator.execute(task, context); const duration = Date.now() - startTime; // 保存到历史 historyManager.save({ task: `[${workflow.name}] ${task}`, summary: result.summary, experts: result.outputs.map(o => o.expertId), outputs: result.outputs.map(o => ({ expertId: o.expertId, expertName: o.expertName, content: o.content, })), conversation: result.conversation.map(m => ({ from: m.from, content: m.content, type: m.type, })), duration, }); const progressText = `\n\n---\n📊 **执行过程**:\n${progressLogs.join('\n')}\n⏱️ 总耗时: ${(duration / 1000).toFixed(1)}s`; return { content: [{ type: 'text', text: `# 🔄 ${workflow.name} 执行结果\n\n${result.summary}${progressText}` }], }; }
• src/server.ts:365-387 (registration)

  Tool registration in the list of available tools, including name, description, and input schema with predefined workflow enum.
  { name: 'run_workflow', description: '使用指定工作流执行任务', inputSchema: { type: 'object', properties: { workflow: { type: 'string', enum: ['code-generation', 'bug-fix', 'refactoring', 'code-review', 'documentation'], description: '工作流 ID', }, task: { type: 'string', description: '任务描述', }, context: { type: 'string', description: '额外上下文（可选）', }, }, required: ['workflow', 'task'], }, },
• src/server.ts:368-387 (schema)

  Input schema definition for the run_workflow tool, specifying required workflow ID (enum), task, and optional context.
  inputSchema: { type: 'object', properties: { workflow: { type: 'string', enum: ['code-generation', 'bug-fix', 'refactoring', 'code-review', 'documentation'], description: '工作流 ID', }, task: { type: 'string', description: '任务描述', }, context: { type: 'string', description: '额外上下文（可选）', }, }, required: ['workflow', 'task'], }, },
• src/collaboration/workflow.ts:71-322 (helper)

  Predefined workflow templates (WORKFLOW_TEMPLATES) that match the enum in run_workflow schema, each defining steps, experts, triggers, etc.
  export const WORKFLOW_TEMPLATES: Record<string, WorkflowDefinition> = { 'code-generation': { id: 'code-generation', name: '代码生成工作流', description: '用于从需求生成代码的标准流程', triggers: ['写', '生成', '创建', '实现', '开发'], entryPoint: 'design', steps: [ { id: 'design', name: '架构设计', type: 'expert', expert: { name: '架构专家', role: '你是一位资深软件架构师。请分析需求，设计技术方案，输出：1) 技术选型 2) 模块划分 3) 接口定义 4) 数据结构', tier: 'powerful', skills: ['架构设计', '系统分析'], }, taskTemplate: '请为以下需求设计技术方案：{{task}}', }, { id: 'implement', name: '代码实现', type: 'expert', dependencies: ['design'], expert: { name: '开发专家', role: '你是一位资深开发工程师。请根据设计方案实现代码，确保代码质量、可读性和可维护性。', tier: 'balanced', skills: ['编码', '最佳实践'], }, taskTemplate: '请根据以上设计方案实现代码', }, { id: 'test', name: '测试用例', type: 'expert', dependencies: ['implement'], expert: { name: '测试专家', role: '你是一位资深测试工程师。请为代码编写单元测试和集成测试，覆盖主要功能和边界情况。', tier: 'balanced', skills: ['测试', '质量保证'], }, taskTemplate: '请为以上代码编写测试用例', optional: true, }, { id: 'review', name: '代码审查', type: 'review', dependencies: ['implement'], }, ], }, 'bug-fix': { id: 'bug-fix', name: 'Bug 修复工作流', description: '用于诊断和修复 Bug 的标准流程', triggers: ['修复', 'bug', '错误', '问题', 'fix', 'error'], entryPoint: 'diagnose', steps: [ { id: 'diagnose', name: '问题诊断', type: 'expert', expert: { name: '诊断专家', role: '你是一位 Bug 诊断专家。请分析错误信息和代码，找出 Bug 的根因，输出：1) 问题描述 2) 根因分析 3) 影响范围', tier: 'powerful', skills: ['调试', '问题分析'], }, taskTemplate: '请诊断以下问题：{{task}}', }, { id: 'fix', name: '修复实现', type: 'expert', dependencies: ['diagnose'], expert: { name: '修复专家', role: '你是一位 Bug 修复专家。请根据诊断结果修复问题，确保修复方案不引入新问题。', tier: 'balanced', skills: ['修复', '重构'], }, taskTemplate: '请根据诊断结果修复 Bug', retries: 2, }, { id: 'verify', name: '验证修复', type: 'expert', dependencies: ['fix'], expert: { name: '验证专家', role: '你是一位 QA 专家。请验证修复是否有效，检查是否引入回归问题。', tier: 'fast', skills: ['验证', '测试'], }, taskTemplate: '请验证以上修复是否正确', }, ], }, 'refactoring': { id: 'refactoring', name: '代码重构工作流', description: '用于代码重构的标准流程', triggers: ['重构', '优化', '改进', 'refactor', 'optimize'], entryPoint: 'analyze', steps: [ { id: 'analyze', name: '代码分析', type: 'expert', expert: { name: '分析专家', role: '你是一位代码质量分析专家。请分析代码存在的问题：1) 代码坏味道 2) 性能瓶颈 3) 可维护性问题 4) 安全隐患', tier: 'powerful', skills: ['代码分析', '质量评估'], }, taskTemplate: '请分析以下代码的问题：{{task}}', }, { id: 'plan', name: '重构计划', type: 'expert', dependencies: ['analyze'], expert: { name: '重构规划师', role: '你是一位重构专家。请制定详细的重构计划，确保重构过程安全可控。', tier: 'balanced', skills: ['重构', '规划'], }, taskTemplate: '请根据分析结果制定重构计划', }, { id: 'execute', name: '执行重构', type: 'expert', dependencies: ['plan'], expert: { name: '重构执行者', role: '你是一位资深开发者。请按计划执行重构，每一步都要保证代码可运行。', tier: 'balanced', skills: ['重构', '编码'], }, taskTemplate: '请执行重构计划', }, { id: 'review', name: '重构审查', type: 'review', dependencies: ['execute'], }, ], }, 'code-review': { id: 'code-review', name: '代码审查工作流', description: '多维度代码审查流程', triggers: ['审查', 'review', '检查', 'check'], entryPoint: 'parallel-review', steps: [ { id: 'parallel-review', name: '并行审查', type: 'parallel', parallel: ['security-review', 'quality-review', 'performance-review'], }, { id: 'security-review', name: '安全审查', type: 'expert', expert: { name: '安全专家', role: '你是一位安全专家。请审查代码的安全性：1) SQL 注入 2) XSS 3) 认证授权 4) 敏感数据处理', tier: 'powerful', skills: ['安全', '漏洞分析'], }, taskTemplate: '请从安全角度审查以下代码：{{task}}', }, { id: 'quality-review', name: '质量审查', type: 'expert', expert: { name: '质量专家', role: '你是一位代码质量专家。请审查代码质量：1) 可读性 2) 可维护性 3) 命名规范 4) 注释完整性', tier: 'balanced', skills: ['代码质量', '最佳实践'], }, taskTemplate: '请从质量角度审查以下代码：{{task}}', }, { id: 'performance-review', name: '性能审查', type: 'expert', expert: { name: '性能专家', role: '你是一位性能优化专家。请审查代码性能：1) 算法复杂度 2) 内存使用 3) IO 操作 4) 缓存策略', tier: 'balanced', skills: ['性能', '优化'], }, taskTemplate: '请从性能角度审查以下代码：{{task}}', }, { id: 'aggregate', name: '汇总审查结果', type: 'aggregate', dependencies: ['security-review', 'quality-review', 'performance-review'], }, ], }, 'documentation': { id: 'documentation', name: '文档生成工作流', description: '自动生成技术文档', triggers: ['文档', '说明', 'doc', 'readme', '注释'], entryPoint: 'analyze-code', steps: [ { id: 'analyze-code', name: '代码分析', type: 'expert', expert: { name: '代码分析师', role: '你是一位代码分析专家。请分析代码结构：1) 模块组成 2) 公开接口 3) 依赖关系 4) 使用方式', tier: 'balanced', skills: ['代码分析', '结构理解'], }, taskTemplate: '请分析以下代码的结构：{{task}}', }, { id: 'generate-doc', name: '生成文档', type: 'expert', dependencies: ['analyze-code'], expert: { name: '文档专家', role: '你是一位技术文档专家。请生成清晰、完整的技术文档，包含：1) 概述 2) 安装使用 3) API 说明 4) 示例代码', tier: 'balanced', skills: ['文档写作', '技术沟通'], }, taskTemplate: '请根据分析结果生成技术文档', }, ], }, };
• src/collaboration/workflow.ts:386-431 (helper)

  WorkflowManager.toTaskAnalysis method: converts a workflow definition into experts and subtasks format used by the orchestrator; called in handler for expert planning.
  toTaskAnalysis(workflow: WorkflowDefinition, task: string): { experts: DynamicExpert[]; subtasks: SubTask[]; workflow: WorkflowType; } { const experts: DynamicExpert[] = []; const subtasks: SubTask[] = []; let expertIndex = 0; // 收集所有专家步骤 const expertSteps = workflow.steps.filter(s => s.type === 'expert' && s.expert); for (const step of expertSteps) { if (!step.expert) continue; const expertId = `expert-${expertIndex++}`; experts.push({ id: expertId, name: step.expert.name, role: step.expert.role, tier: step.expert.tier, skills: step.expert.skills || [], }); subtasks.push({ id: step.id, description: step.taskTemplate?.replace('{{task}}', task) || task, expertId, dependencies: step.dependencies || [], priority: expertIndex, }); } // 判断工作流类型 const hasParallel = workflow.steps.some(s => s.type === 'parallel'); const hasDependencies = workflow.steps.some(s => s.dependencies && s.dependencies.length > 0); let workflowType: WorkflowType = 'sequential'; if (hasParallel && !hasDependencies) { workflowType = 'parallel'; } else if (hasParallel && hasDependencies) { workflowType = 'mixed'; } return { experts, subtasks, workflow: workflowType }; }