New Hook Builder /new-hook

Create and configure Claude Code hooks for automating workflows, notifications, formatting, and behavior control.

Arguments

• [hook-description] - Description of the hook you want to create

Example Usage

# Create a logging hook
/new-hook "log all bash commands to a file"

# Create a protection hook
/new-hook "prevent editing .env files"

# Create a formatting hook
/new-hook "auto-format TypeScript files after editing"

# Create a notification hook
/new-hook "send desktop notification when Claude needs input"

Hook Events Overview

Claude Code provides these hook events:

• PreToolUse: Runs before tool calls (can block them with exit codes)
• PostToolUse: Runs after tool calls complete
• UserPromptSubmit: Runs when user submits a prompt, before Claude processes it
• Notification: Runs when Claude Code sends notifications
• Stop: Runs when Claude Code finishes responding
• SubagentStop: Runs when subagent tasks complete
• PreCompact: Runs before Claude Code runs a compact operation
• SessionStart: Runs when Claude Code starts/resumes a session

Documentation References

For comprehensive information, refer to:

• Claude Code Hooks Guide - Complete documentation and examples
• Claude Code Hooks Reference - Technical reference and API details

Instructions

When creating hooks, follow these steps:

1. Analyze Requirements: Understand what the user wants to automate or control
2. Select Hook Event: Choose the appropriate event type based on when the action should occur
3. Design Matcher: Create matcher patterns to target specific tools (use * for all tools)
4. Write Hook Command: Create the shell command that will execute
5. Consider Security: Review security implications and add safeguards
6. Choose Storage Location: Use AskUserQuestion to ask: user settings (global) or project settings (local)?
7. Generate Configuration: Create proper JSON configuration
8. Test Hook: Verify the hook works as expected

Common Hook Patterns

Tool Matchers:

• Bash - Shell commands only
• Edit|MultiEdit|Write - File editing operations
• Read - File reading operations
• * - All tools

Exit Codes (PreToolUse only):

• 0 - Allow tool execution
• 2 - Block tool execution with feedback
• Other non-zero - Block silently

Security Considerations

⚠️ CRITICAL: Hooks run automatically with your environment's credentials. Always:

• Review hook commands before registering
• Validate file paths and inputs
• Use absolute paths when possible
• Avoid executing untrusted code
• Test hooks in safe environments first

Configuration Format

Hooks are stored in JSON format in ~/.claude/settings.json (user) or <project>/.claude/settings.json (project):

{
  "hooks": {
    "<EventType>": [
      {
        "matcher": "<tool-pattern>",
        "hooks": [
          {
            "type": "command",
            "command": "<shell-command>"
          }
        ]
      }
    ]
  }
}

Common Examples

Command Logging (PreToolUse):

jq -r '"\(.tool_input.command) - \(.tool_input.description // "No description")"' >> ~/.claude/command-log.txt

Code Formatting (PostToolUse):

jq -r '.tool_input.file_path' | { read file_path; if echo "$file_path" | grep -q '\.ts$'; then npx prettier --write "$file_path"; fi; }

File Protection (PreToolUse):

python3 -c "import json, sys; data=json.load(sys.stdin); path=data.get('tool_input',{}).get('file_path',''); sys.exit(2 if any(p in path for p in ['.env', 'package-lock.json', '.git/']) else 0)"

Desktop Notifications (Notification):

notify-send 'Claude Code' 'Awaiting your input'

Best Practices

• Start with simple hooks and iterate
• Use descriptive matcher patterns
• Include error handling in hook commands
• Test hooks thoroughly before deployment
• Document complex hook logic
• Use project settings for team-shared hooks
• Use user settings for personal preferences
• Monitor hook performance and logs
• Keep hook commands concise and focused

Response Format

After creating a hook, provide:

1. Hook Configuration: Complete JSON configuration
2. Installation Instructions: Where to add the configuration
3. Testing Steps: How to verify the hook works
4. Security Review: Any security considerations
5. Usage Examples: How the hook will behave in practice

Always explain the hook's purpose, when it triggers, and what it accomplishes.

Additional Resources

• Use the /hooks slash command in Claude Code to configure hooks interactively
• Check existing hook configurations in ~/.claude/settings.json or <project>/.claude/settings.json
• Reference the official Claude Code hooks documentation for the latest examples and best practices