Hooks Automation

Hooks 自动化

通过与 MCP 工具集成的钩子和神经模式训练，智能自动化系统可协调、验证 Claude Code 操作，并从中学习。

此技能的功能

此技能提供一套全面的钩子系统，可自动管理开发操作、协调 swarm Agent、维护会话状态，并持续从编码模式中学习。它支持自动分配 Agent、代码格式化、性能跟踪和跨会话记忆持久化。

主要功能：

• 操作前钩子：在操作前进行验证、准备并自动分配 Agent
• 操作后钩子：在操作后进行格式化、分析并训练模式
• 会话管理：持久化状态、恢复上下文、生成摘要
• 记忆协调：在 swarm Agent 之间同步知识
• Git 集成：自动化提交钩子并进行质量验证
• 神经训练：从成功模式中持续学习
• MCP 集成：与 swarm 工具无缝协调

前置条件

必需：

• 已安装 Claude Flow CLI（npm install -g claude-flow@alpha）
• 已启用钩子的 Claude Code
• 包含钩子配置的 .claude/settings.json

可选：

• 已配置的 MCP 服务（claude-flow、ruv-swarm、flow-nexus）
• 用于版本控制的 Git 仓库
• 用于质量验证的测试框架

快速开始

初始化钩子系统

# Initialize with default hooks configuration
npx claude-flow init --hooks

这将创建：

• 包含预配置钩子的 .claude/settings.json
• .claude/commands/hooks/ 中的钩子命令文档
• 用于常见操作的默认钩子处理程序

基本钩子用法

# Pre-task hook (auto-spawns agents)
npx claude-flow hook pre-task --description "Implement authentication"

# Post-edit hook (auto-formats and stores in memory)
npx claude-flow hook post-edit --file "src/auth.js" --memory-key "auth/login"

# Session end hook (saves state and metrics)
npx claude-flow hook session-end --session-id "dev-session" --export-metrics

---

完整指南

可用钩子

操作前钩子

这些钩子会在操作前执行，以进行准备和验证：

pre-edit - 在修改文件前进行验证并分配 Agent

npx claude-flow hook pre-edit [options]

Options:
  --file, -f <path>         File path to be edited
  --auto-assign-agent       Automatically assign best agent (default: true)
  --validate-syntax         Pre-validate syntax before edit
  --check-conflicts         Check for merge conflicts
  --backup-file             Create backup before editing

Examples:
  npx claude-flow hook pre-edit --file "src/auth/login.js"
  npx claude-flow hook pre-edit -f "config/db.js" --validate-syntax
  npx claude-flow hook pre-edit -f "production.env" --backup-file --check-conflicts

功能：

• 根据文件类型自动分配 Agent
• 进行语法验证，防止代码损坏
• 检测并发编辑冲突
• 自动备份文件以确保安全

pre-bash - 检查命令安全性和资源需求

npx claude-flow hook pre-bash --command <cmd>

Options:
  --command, -c <cmd>       Command to validate
  --check-safety            Verify command safety (default: true)
  --estimate-resources      Estimate resource usage
  --require-confirmation    Request user confirmation for risky commands

Examples:
  npx claude-flow hook pre-bash -c "rm -rf /tmp/cache"
  npx claude-flow hook pre-bash --command "docker build ." --estimate-resources

功能：

• 命令安全性验证
• 资源需求估算
• 破坏性命令确认
• 权限检查

pre-task - 自动启动 Agent 并为复杂任务做好准备

npx claude-flow hook pre-task [options]

Options:
  --description, -d <text>  Task description for context
  --auto-spawn-agents       Automatically spawn required agents (default: true)
  --load-memory             Load relevant memory from previous sessions
  --optimize-topology       Select optimal swarm topology
  --estimate-complexity     Analyze task complexity

Examples:
  npx claude-flow hook pre-task --description "Implement user authentication"
  npx claude-flow hook pre-task -d "Continue API dev" --load-memory
  npx claude-flow hook pre-task -d "Refactor codebase" --optimize-topology

功能：

• 根据任务分析自动启动 Agent
• 加载记忆以保持上下文连贯
• 针对任务结构优化拓扑
• 估算复杂度并预测耗时

pre-search - 准备并优化搜索操作

npx claude-flow hook pre-search --query <query>

Options:
  --query, -q <text>        Search query
  --check-cache             Check cache first (default: true)
  --optimize-query          Optimize search pattern

Examples:
  npx claude-flow hook pre-search -q "authentication middleware"

功能：

• 检查缓存以加快结果获取
• 优化查询
• 改进搜索模式

操作后钩子

这些钩子会在操作后执行，以进行处理和学习：

post-edit - 自动格式化、验证并更新记忆

npx claude-flow hook post-edit [options]

Options:
  --file, -f <path>         File path that was edited
  --auto-format             Automatically format code (default: true)
  --memory-key, -m <key>    Store edit context in memory
  --train-patterns          Train neural patterns from edit
  --validate-output         Validate edited file

Examples:
  npx claude-flow hook post-edit --file "src/components/Button.jsx"
  npx claude-flow hook post-edit -f "api/auth.js" --memory-key "auth/login"
  npx claude-flow hook post-edit -f "utils/helpers.ts" --train-patterns

功能：

• 针对特定语言自动格式化（Prettier、Black、gofmt）
• 存储编辑上下文和决策的记忆
• 通过神经模式训练持续改进
• 使用 linting 验证输出

post-bash - 记录执行情况并更新指标

npx claude-flow hook post-bash --command <cmd>

Options:
  --command, -c <cmd>       Command that was executed
  --log-output              Log command output (default: true)
  --update-metrics          Update performance metrics
  --store-result            Store result in memory

Examples:
  npx claude-flow hook post-bash -c "npm test" --update-metrics

功能：

• 记录命令执行情况
• 跟踪性能指标
• 存储结果以供分析
• 检测错误模式

post-task - 分析性能并存储决策

npx claude-flow hook post-task [options]

Options:
  --task-id, -t <id>        Task identifier for tracking
  --analyze-performance     Generate performance metrics (default: true)
  --store-decisions         Save task decisions to memory
  --export-learnings        Export neural pattern learnings
  --generate-report         Create task completion report

Examples:
  npx claude-flow hook post-task --task-id "auth-implementation"
  npx claude-flow hook post-task -t "api-refactor" --analyze-performance
  npx claude-flow hook post-task -t "bug-fix-123" --store-decisions

功能：

• 测量执行时间和 token 用量
• 记录决策和实现方案的选择
• 导出神经学习模式
• 生成完成报告

post-search - 缓存结果并改进模式

npx claude-flow hook post-search --query <query> --results <path>

Options:
  --query, -q <text>        Original search query
  --results, -r <path>      Results file path
  --cache-results           Cache for future use (default: true)
  --train-patterns          Improve search patterns

Examples:
  npx claude-flow hook post-search -q "auth" -r "results.json" --train-patterns

功能：

• 缓存结果以加快后续搜索
• 改进搜索模式
• 评估相关性

MCP 集成钩子

这些钩子会与 MCP swarm 工具协调：

mcp-initialized - 持久化 swarm 配置

npx claude-flow hook mcp-initialized --swarm-id <id>

Features:
- Save swarm topology and configuration
- Store agent roster in memory
- Initialize coordination namespace

agent-spawned - 更新 Agent 名单和记忆

npx claude-flow hook agent-spawned --agent-id <id> --type <type>

Features:
- Register agent in coordination memory
- Update agent roster
- Initialize agent-specific memory namespace

task-orchestrated - 监控任务进度

npx claude-flow hook task-orchestrated --task-id <id>

Features:
- Track task progress through memory
- Monitor agent assignments
- Update coordination state

neural-trained - 保存模式改进

npx claude-flow hook neural-trained --pattern <name>

Features:
- Export trained neural patterns
- Update coordination models
- Share learning across agents

记忆协调钩子

memory-write - 在 Agent 写入协调记忆时触发

Features:
- Validate memory key format
- Update cross-agent indexes
- Trigger dependent hooks
- Notify subscribed agents

memory-read - 在 Agent 读取协调记忆时触发

Features:
- Log access patterns
- Update popularity metrics
- Preload related data
- Track usage statistics

memory-sync - 在各个 swarm Agent 之间同步记忆

npx claude-flow hook memory-sync --namespace <ns>

Features:
- Sync memory state across agents
- Resolve conflicts
- Propagate updates
- Maintain consistency

会话钩子

session-start - 初始化新会话

npx claude-flow hook session-start --session-id <id>

Options:
  --session-id, -s <id>     Session identifier
  --load-context            Load context from previous session
  --init-agents             Initialize required agents

Features:
- Create session directory
- Initialize metrics tracking
- Load previous context
- Set up coordination namespace

session-restore - 加载之前的会话状态

npx claude-flow hook session-restore --session-id <id>

Options:
  --session-id, -s <id>     Session to restore
  --restore-memory          Restore memory state (default: true)
  --restore-agents          Restore agent configurations

Examples:
  npx claude-flow hook session-restore --session-id "swarm-20241019"
  npx claude-flow hook session-restore -s "feature-auth" --restore-memory

功能：

• 加载之前的会话上下文
• 恢复记忆状态和决策
• 将 Agent 重新配置为之前的状态
• 恢复进行中的任务

session-end - 清理并持久化会话状态

npx claude-flow hook session-end [options]

Options:
  --session-id, -s <id>     Session identifier to end
  --save-state              Save current session state (default: true)
  --export-metrics          Export session metrics
  --generate-summary        Create session summary
  --cleanup-temp            Remove temporary files

Examples:
  npx claude-flow hook session-end --session-id "dev-session-2024"
  npx claude-flow hook session-end -s "feature-auth" --export-metrics --generate-summary
  npx claude-flow hook session-end -s "quick-fix" --cleanup-temp

功能：

• 保存当前上下文和进度
• 导出会话指标（时长、命令、令牌、文件）
• 生成包含决策和后续步骤的工作摘要
• 清理临时文件并优化存储

notify - 发送包含 swarm 状态的自定义通知

npx claude-flow hook notify --message <msg>

Options:
  --message, -m <text>      Notification message
  --level <level>           Notification level (info|warning|error)
  --swarm-status            Include swarm status (default: true)
  --broadcast               Send to all agents

Examples:
  npx claude-flow hook notify -m "Task completed" --level info
  npx claude-flow hook notify -m "Critical error" --level error --broadcast

功能：

• 向协调系统发送通知
• 包含 swarm 状态和指标
• 向所有 Agent 广播
• 记录重要事件

配置

基本配置

编辑 .claude/settings.json 以配置钩子：

{
  "hooks": {
    "PreToolUse": [
      {
        "matcher": "^(Write|Edit|MultiEdit)$",
        "hooks": [{
          "type": "command",
          "command": "npx claude-flow hook pre-edit --file '${tool.params.file_path}' --memory-key 'swarm/editor/current'"
        }]
      },
      {
        "matcher": "^Bash$",
        "hooks": [{
          "type": "command",
          "command": "npx claude-flow hook pre-bash --command '${tool.params.command}'"
        }]
      }
    ],
    "PostToolUse": [
      {
        "matcher": "^(Write|Edit|MultiEdit)$",
        "hooks": [{
          "type": "command",
          "command": "npx claude-flow hook post-edit --file '${tool.params.file_path}' --memory-key 'swarm/editor/complete' --auto-format --train-patterns"
        }]
      },
      {
        "matcher": "^Bash$",
        "hooks": [{
          "type": "command",
          "command": "npx claude-flow hook post-bash --command '${tool.params.command}' --update-metrics"
        }]
      }
    ]
  }
}

高级配置

包含所有功能的完整钩子配置：

{
  "hooks": {
    "enabled": true,
    "debug": false,
    "timeout": 5000,

    "PreToolUse": [
      {
        "matcher": "^(Write|Edit|MultiEdit)$",
        "hooks": [
          {
            "type": "command",
            "command": "npx claude-flow hook pre-edit --file '${tool.params.file_path}' --auto-assign-agent --validate-syntax",
            "timeout": 3000,
            "continueOnError": true
          }
        ]
      },
      {
        "matcher": "^Task$",
        "hooks": [
          {
            "type": "command",
            "command": "npx claude-flow hook pre-task --description '${tool.params.task}' --auto-spawn-agents --load-memory",
            "async": true
          }
        ]
      },
      {
        "matcher": "^Grep$",
        "hooks": [
          {
            "type": "command",
            "command": "npx claude-flow hook pre-search --query '${tool.params.pattern}' --check-cache"
          }
        ]
      }
    ],

    "PostToolUse": [
      {
        "matcher": "^(Write|Edit|MultiEdit)$",
        "hooks": [
          {
            "type": "command",
            "command": "npx claude-flow hook post-edit --file '${tool.params.file_path}' --memory-key 'edits/${tool.params.file_path}' --auto-format --train-patterns",
            "async": true
          }
        ]
      },
      {
        "matcher": "^Task$",
        "hooks": [
          {
            "type": "command",
            "command": "npx claude-flow hook post-task --task-id '${result.task_id}' --analyze-performance --store-decisions --export-learnings",
            "async": true
          }
        ]
      },
      {
        "matcher": "^Grep$",
        "hooks": [
          {
            "type": "command",
            "command": "npx claude-flow hook post-search --query '${tool.params.pattern}' --cache-results --train-patterns"
          }
        ]
      }
    ],

    "SessionStart": [
      {
        "hooks": [
          {
            "type": "command",
            "command": "npx claude-flow hook session-start --session-id '${session.id}' --load-context"
          }
        ]
      }
    ],

    "SessionEnd": [
      {
        "hooks": [
          {
            "type": "command",
            "command": "npx claude-flow hook session-end --session-id '${session.id}' --export-metrics --generate-summary --cleanup-temp"
          }
        ]
      }
    ]
  }
}

受保护的文件模式

为敏感文件添加保护：

{
  "hooks": {
    "PreToolUse": [
      {
        "matcher": "^(Write|Edit|MultiEdit)$",
        "hooks": [
          {
            "type": "command",
            "command": "npx claude-flow hook check-protected --file '${tool.params.file_path}'"
          }
        ]
      }
    ]
  }
}

自动测试

在文件修改后运行测试：

{
  "hooks": {
    "PostToolUse": [
      {
        "matcher": "^Write$",
        "hooks": [
          {
            "type": "command",
            "command": "test -f '${tool.params.file_path%.js}.test.js' && npm test '${tool.params.file_path%.js}.test.js'",
            "continueOnError": true
          }
        ]
      }
    ]
  }
}

MCP 工具集成

钩子会自动与 MCP 工具集成以实现协调：

生成 Agent 的任务前钩子

// Hook command
npx claude-flow hook pre-task --description "Build REST API"

// Internally calls MCP tools:
mcp__claude-flow__agent_spawn {
  type: "backend-dev",
  capabilities: ["api", "database", "testing"]
}

mcp__claude-flow__memory_usage {
  action: "store",
  key: "swarm/task/api-build/context",
  namespace: "coordination",
  value: JSON.stringify({
    description: "Build REST API",
    agents: ["backend-dev"],
    started: Date.now()
  })
}

存储记忆的编辑后钩子

// Hook command
npx claude-flow hook post-edit --file "api/auth.js"

// Internally calls MCP tools:
mcp__claude-flow__memory_usage {
  action: "store",
  key: "swarm/edits/api/auth.js",
  namespace: "coordination",
  value: JSON.stringify({
    file: "api/auth.js",
    timestamp: Date.now(),
    changes: { added: 45, removed: 12 },
    formatted: true,
    linted: true
  })
}

mcp__claude-flow__neural_train {
  pattern_type: "coordination",
  training_data: { /* edit patterns */ }
}

持久化状态的会话结束钩子

// Hook command
npx claude-flow hook session-end --session-id "dev-2024"

// Internally calls MCP tools:
mcp__claude-flow__memory_persist {
  sessionId: "dev-2024"
}

mcp__claude-flow__swarm_status {
  swarmId: "current"
}

// Generates metrics and summary

记忆协调协议

所有钩子都遵循标准化的记忆协调模式：

三阶段记忆协议

阶段 1：STATUS - 钩子启动

mcp__claude-flow__memory_usage {
  action: "store",
  key: "swarm/hooks/pre-edit/status",
  namespace: "coordination",
  value: JSON.stringify({
    status: "running",
    hook: "pre-edit",
    file: "src/auth.js",
    timestamp: Date.now()
  })
}

阶段 2：PROGRESS - 钩子处理

mcp__claude-flow__memory_usage {
  action: "store",
  key: "swarm/hooks/pre-edit/progress",
  namespace: "coordination",
  value: JSON.stringify({
    progress: 50,
    action: "validating syntax",
    file: "src/auth.js"
  })
}

阶段 3：COMPLETE - 钩子完成

mcp__claude-flow__memory_usage {
  action: "store",
  key: "swarm/hooks/pre-edit/complete",
  namespace: "coordination",
  value: JSON.stringify({
    status: "complete",
    result: "success",
    agent_assigned: "backend-dev",
    syntax_valid: true,
    backup_created: true
  })
}

钩子响应格式

钩子会返回 JSON 响应以控制操作流程：

继续响应

{
  "continue": true,
  "reason": "All validations passed",
  "metadata": {
    "agent_assigned": "backend-dev",
    "syntax_valid": true,
    "file": "src/auth.js"
  }
}

阻止响应

{
  "continue": false,
  "reason": "Protected file - manual review required",
  "metadata": {
    "file": ".env.production",
    "protection_level": "high",
    "requires": "manual_approval"
  }
}

警告响应

{
  "continue": true,
  "reason": "Syntax valid but complexity high",
  "warnings": [
    "Cyclomatic complexity: 15 (threshold: 10)",
    "Consider refactoring for better maintainability"
  ],
  "metadata": {
    "complexity": 15,
    "threshold": 10
  }
}

Git 集成

钩子可以与 Git 操作集成以实现质量控制：

提交前钩子

# Add to .git/hooks/pre-commit or use husky

#!/bin/bash
# Run quality checks before commit

# Get staged files
FILES=$(git diff --cached --name-only --diff-filter=ACM)

for FILE in $FILES; do
  # Run pre-edit hook for validation
  npx claude-flow hook pre-edit --file "$FILE" --validate-syntax

  if [ $? -ne 0 ]; then
    echo "Validation failed for $FILE"
    exit 1
  fi

  # Run post-edit hook for formatting
  npx claude-flow hook post-edit --file "$FILE" --auto-format
done

# Run tests
npm test

exit $?

提交后钩子

# Add to .git/hooks/post-commit

#!/bin/bash
# Track commit metrics

COMMIT_HASH=$(git rev-parse HEAD)
COMMIT_MSG=$(git log -1 --pretty=%B)

npx claude-flow hook notify \
  --message "Commit completed: $COMMIT_MSG" \
  --level info \
  --swarm-status

推送前钩子

# Add to .git/hooks/pre-push

#!/bin/bash
# Quality gate before push

# Run full test suite
npm run test:all

# Run quality checks
npx claude-flow hook session-end \
  --generate-report \
  --export-metrics

# Verify quality thresholds
TRUTH_SCORE=$(npx claude-flow metrics score --format json | jq -r '.truth_score')

if (( $(echo "$TRUTH_SCORE < 0.95" | bc -l) )); then
  echo "Truth score below threshold: $TRUTH_SCORE < 0.95"
  exit 1
fi

exit 0

Agent 协调工作流

Agent 如何使用钩子进行协调：

Agent 工作流示例

# Agent 1: Backend Developer
# STEP 1: Pre-task preparation
npx claude-flow hook pre-task \
  --description "Implement user authentication API" \
  --auto-spawn-agents \
  --load-memory

# STEP 2: Work begins - pre-edit validation
npx claude-flow hook pre-edit \
  --file "api/auth.js" \
  --auto-assign-agent \
  --validate-syntax

# STEP 3: Edit file (via Claude Code Edit tool)
# ... code changes ...

# STEP 4: Post-edit processing
npx claude-flow hook post-edit \
  --file "api/auth.js" \
  --memory-key "swarm/backend/auth-api" \
  --auto-format \
  --train-patterns

# STEP 5: Notify coordination system
npx claude-flow hook notify \
  --message "Auth API implementation complete" \
  --swarm-status \
  --broadcast

# STEP 6: Task completion
npx claude-flow hook post-task \
  --task-id "auth-api" \
  --analyze-performance \
  --store-decisions \
  --export-learnings

# Agent 2: Test Engineer (receives notification)
# STEP 1: Check memory for API details
npx claude-flow hook session-restore \
  --session-id "swarm-current" \
  --restore-memory

# Memory contains: swarm/backend/auth-api with implementation details

# STEP 2: Generate tests
npx claude-flow hook pre-task \
  --description "Write tests for auth API" \
  --load-memory

# STEP 3: Create test file
npx claude-flow hook post-edit \
  --file "api/auth.test.js" \
  --memory-key "swarm/testing/auth-api-tests" \
  --train-patterns

# STEP 4: Share test results
npx claude-flow hook notify \
  --message "Auth API tests complete - 100% coverage" \
  --broadcast

自定义钩子创建

为特定工作流创建自定义钩子：

自定义钩子模板

// .claude/hooks/custom-quality-check.js

module.exports = {
  name: 'custom-quality-check',
  type: 'pre',
  matcher: /\.(ts|js)$/,

  async execute(context) {
    const { file, content } = context;

    // Custom validation logic
    const complexity = await analyzeComplexity(content);
    const securityIssues = await scanSecurity(content);

    // Store in memory
    await storeInMemory({
      key: `quality/${file}`,
      value: { complexity, securityIssues }
    });

    // Return decision
    if (complexity > 15 || securityIssues.length > 0) {
      return {
        continue: false,
        reason: 'Quality checks failed',
        warnings: [
          `Complexity: ${complexity} (max: 15)`,
          `Security issues: ${securityIssues.length}`
        ]
      };
    }

    return {
      continue: true,
      reason: 'Quality checks passed',
      metadata: { complexity, securityIssues: 0 }
    };
  }
};

注册自定义钩子

{
  "hooks": {
    "PreToolUse": [
      {
        "matcher": "^(Write|Edit)$",
        "hooks": [
          {
            "type": "script",
            "script": ".claude/hooks/custom-quality-check.js"
          }
        ]
      }
    ]
  }
}

实际示例

示例 1：全栈开发工作流

# Session start - initialize coordination
npx claude-flow hook session-start --session-id "fullstack-feature"

# Pre-task planning
npx claude-flow hook pre-task \
  --description "Build user profile feature - frontend + backend + tests" \
  --auto-spawn-agents \
  --optimize-topology

# Backend work
npx claude-flow hook pre-edit --file "api/profile.js"
# ... implement backend ...
npx claude-flow hook post-edit \
  --file "api/profile.js" \
  --memory-key "profile/backend" \
  --train-patterns

# Frontend work (reads backend details from memory)
npx claude-flow hook pre-edit --file "components/Profile.jsx"
# ... implement frontend ...
npx claude-flow hook post-edit \
  --file "components/Profile.jsx" \
  --memory-key "profile/frontend" \
  --train-patterns

# Testing (reads both backend and frontend from memory)
npx claude-flow hook pre-task \
  --description "Test profile feature" \
  --load-memory

# Session end - export everything
npx claude-flow hook session-end \
  --session-id "fullstack-feature" \
  --export-metrics \
  --generate-summary

示例 2：使用钩子进行调试

# Start debugging session
npx claude-flow hook session-start --session-id "debug-memory-leak"

# Pre-task: analyze issue
npx claude-flow hook pre-task \
  --description "Debug memory leak in event handlers" \
  --load-memory \
  --estimate-complexity

# Search for event emitters
npx claude-flow hook pre-search --query "EventEmitter"
# ... search executes ...
npx claude-flow hook post-search \
  --query "EventEmitter" \
  --cache-results

# Fix the issue
npx claude-flow hook pre-edit \
  --file "services/events.js" \
  --backup-file
# ... fix code ...
npx claude-flow hook post-edit \
  --file "services/events.js" \
  --memory-key "debug/memory-leak-fix" \
  --validate-output

# Verify fix
npx claude-flow hook post-task \
  --task-id "memory-leak-fix" \
  --analyze-performance \
  --generate-report

# End session
npx claude-flow hook session-end \
  --session-id "debug-memory-leak" \
  --export-metrics

示例 3：多 Agent 重构

# Initialize swarm for refactoring
npx claude-flow hook pre-task \
  --description "Refactor legacy codebase to modern patterns" \
  --auto-spawn-agents \
  --optimize-topology

# Agent 1: Code Analyzer
npx claude-flow hook pre-task --description "Analyze code complexity"
# ... analysis ...
npx claude-flow hook post-task \
  --task-id "analysis" \
  --store-decisions

# Agent 2: Refactoring (reads analysis from memory)
npx claude-flow hook session-restore \
  --session-id "swarm-refactor" \
  --restore-memory

for file in src/**/*.js; do
  npx claude-flow hook pre-edit --file "$file" --backup-file
  # ... refactor ...
  npx claude-flow hook post-edit \
    --file "$file" \
    --memory-key "refactor/$file" \
    --auto-format \
    --train-patterns
done

# Agent 3: Testing (reads refactored code from memory)
npx claude-flow hook pre-task \
  --description "Generate tests for refactored code" \
  --load-memory

# Broadcast completion
npx claude-flow hook notify \
  --message "Refactoring complete - all tests passing" \
  --broadcast

性能提示

1. 保持钩子轻量 - 目标执行时间 < 100ms
2. 对耗时操作使用异步处理 - 不要阻塞主流程
3. 积极使用缓存 - 存储频繁访问的数据
4. 批量处理相关操作 - 合并多个操作
5. 合理使用记忆 - 设置恰当的 TTL
6. 监控钩子性能 - 跟踪执行时间
7. 尽可能并行处理 - 并发运行彼此独立的钩子

调试钩子

启用调试模式以排查问题：

# Enable debug output
export CLAUDE_FLOW_DEBUG=true

# Test specific hook with verbose output
npx claude-flow hook pre-edit --file "test.js" --debug

# Check hook execution logs
cat .claude-flow/logs/hooks-$(date +%Y-%m-%d).log

# Validate configuration
npx claude-flow hook validate-config

优势

• 自动分配 Agent：为每种文件类型分配合适的 Agent
• 统一的代码格式：使用特定语言的格式化工具
• 持续学习：神经模式会随时间不断改进
• 跨会话记忆：上下文会在不同会话间持久保留
• 性能跟踪：提供全面的指标和分析
• 自动协调：Agent 通过记忆同步
• 智能生成 Agent：根据任务选择 Agent
• 质量关卡：提交前验证与检查
• 错误预防：编辑前验证语法
• 知识共享：保存并共享决策
• 减少手动工作：自动执行重复性任务
• 加强协作：实现无缝的多 Agent 协调

最佳实践

1. 尽早配置钩子 - 在项目初始化期间设置
2. 策略性地使用记忆键 - 使用清晰的命名空间进行组织
3. 启用自动格式化 - 保持代码一致性
4. 持续训练模式 - 从成功操作中学习
5. 监控性能 - 跟踪钩子执行时间
6. 验证配置 - 在生产环境中使用前测试钩子
7. 记录自定义钩子 - 维护钩子文档
8. 设置合适的超时时间 - 防止操作挂起
9. 妥善处理错误 - 在适当情况下使用 continueOnError
10. 定期查看指标 - 根据使用模式进行优化

故障排除

钩子未执行

• 验证 .claude/settings.json 语法
• 检查钩子匹配器模式
• 启用调试模式
• 查看权限设置
• 确保 claude-flow CLI 位于 PATH 中

钩子超时

• 在配置中增加超时值
• 对耗时操作使用异步钩子
• 优化钩子逻辑
• 检查 MCP 工具的网络连接

记忆问题

• 为记忆键设置合适的 TTL
• 清理旧的记忆条目
• 有效使用记忆命名空间
• 监控记忆用量

性能问题

• 分析钩子执行时间
• 对重复操作使用缓存
• 尽可能批量处理操作
• 降低钩子复杂度

相关命令

• npx claude-flow init --hooks - 初始化钩子系统
• npx claude-flow hook --list - 列出可用钩子
• npx claude-flow hook --test <hook> - 测试指定钩子
• npx claude-flow memory usage - 管理记忆
• npx claude-flow agent spawn - 启动 Agent
• npx claude-flow swarm init - 初始化群组

与其他技能的集成

此技能可与以下内容无缝协作：

• SPARC Methodology - 钩子可增强 SPARC 工作流
• Pair Programming - 在结对编程会话中实现自动化质量保障
• Verification Quality - 在钩子中进行真值评分验证
• GitHub Workflows - 用于提交和 PR 的 Git 集成
• Performance Analysis - 在钩子中收集指标
• Swarm Advanced - 通过钩子实现多 Agent 协作