claude-dev-toolkit

Guided hook creation workflow. Helps you create Claude Code hooks by walking through event selection, matcher configuration, and handler implementation.

Interactive wizard to scaffold a new Claude Code plugin. Creates the complete directory structure, plugin.json, and starter components based on your requirements.

General Q&A about Claude Code extensibility. Ask about skills, plugins, hooks, subagents, teams, MCP servers, output styles, or any Claude Code extension mechanism. Usage - /claude-code-guide [your question]

Use this agent when the user wants to create custom subagents or configure agent teams for Claude Code. Examples: <example>Context: User wants to create a subagent. user: "Create a code review agent that checks for security issues" assistant: "I'll use the agent-builder agent to create a custom subagent with the right tools and system prompt." <commentary>User wants a new custom subagent — the agent-builder handles agent creation from requirements to configuration.</commentary></example> <example>Context: User wants to set up an agent team. user: "I want to set up a team of agents for my project — a researcher, a coder, and a reviewer" assistant: "I'll use the agent-builder agent to design and configure this agent team." <commentary>User needs an agent team — the agent-builder handles team design and agent configuration.</commentary></example> <example>Context: User needs help with agent configuration. user: "My custom agent doesn't have access to the Bash tool, how do I fix that?" assistant: "I'll use the agent-builder agent to diagnose and fix the agent's tool configuration." <commentary>Agent configuration debugging is a core agent-builder capability.</commentary></example>

Use this agent when the user wants to create, configure, or debug Claude Code hooks. Examples: <example>Context: User wants to create a new hook. user: "I want to block rm -rf commands in Claude Code" assistant: "I'll use the hook-builder agent to create a PreToolUse hook that blocks dangerous commands." <commentary>User wants to create a safety hook — the hook-builder handles hook creation from requirements to configuration.</commentary></example> <example>Context: User wants notifications. user: "Can Claude send me a desktop notification when it finishes a task?" assistant: "I'll use the hook-builder agent to create a Stop hook with desktop notifications." <commentary>Desktop notifications on task completion is a classic hook use case.</commentary></example> <example>Context: User's hook isn't working. user: "My PostToolUse hook isn't firing when I edit files" assistant: "I'll use the hook-builder agent to diagnose and fix the hook configuration." <commentary>Hook debugging is a core hook-builder capability.</commentary></example>

Use this agent when the user wants to create a new Claude Code plugin from scratch, scaffold plugin structure, or validate an existing plugin's structure and conventions. Examples: <example>Context: User wants to create a new plugin. user: "Help me build a Claude Code plugin for code review automation" assistant: "I'll use the plugin-builder agent to scaffold and build this plugin for you." <commentary>User wants to create a new plugin from scratch — the plugin-builder agent handles the complete scaffold-to-validation workflow.</commentary></example> <example>Context: User has an existing plugin and wants to validate it. user: "Can you check if my plugin structure is correct?" assistant: "I'll use the plugin-builder agent to validate your plugin's structure and conventions." <commentary>User needs plugin validation, which is part of the plugin-builder's workflow.</commentary></example> <example>Context: User wants to add a component to an existing plugin. user: "Add a new skill and command to my analytics plugin" assistant: "I'll use the plugin-builder agent to add those components to your plugin." <commentary>Adding components to an existing plugin is a core plugin-builder task.</commentary></example>

Comprehensive guide for optimizing Claude Code workflows and effectiveness. Use when writing or improving CLAUDE.md files, managing context windows, crafting effective prompts, running parallel sessions, debugging workflow issues, or improving Claude Code productivity. Covers the explore-plan-code workflow, verification strategies, context management, prompt engineering, session management, and scaling with parallel sessions. Trigger phrases: "best practices", "CLAUDE.md", "context management", "context window", "prompt engineering", "parallel sessions", "Claude Code tips", "effective workflow", "explore-plan-code", "verification strategies", "session management", "fan out", "headless mode", "subagents", "course correct", "rewind".

Comprehensive troubleshooting guide for diagnosing and fixing Claude Code issues across installation, runtime, and plugin systems. Use when troubleshooting Claude Code issues, fixing installation problems, resolving runtime errors, debugging plugin issues, or diagnosing configuration problems. Trigger phrases: "troubleshoot", "not working", "error", "installation issue", "permission denied", "plugin not loading", "skill not triggering", "hook not firing", "MCP server error", "WSL issue", "npm error", "command not found", "Claude Code problem", "debug", "sandbox error", "socat", "bubblewrap", "authentication issue", "API key", "context window", "search not working", "IDE not detected", "JetBrains", "ripgrep", "doctor", "reset config", "high CPU", "memory usage", "freeze", "hang", "PATH issue", "nvm conflict", "node not found", "Git Bash".

Comprehensive guide for extending Claude Code with custom functionality. Use when deciding how to extend Claude Code, comparing extension mechanisms, creating output styles, or running Claude Code programmatically. Trigger phrases: "extend Claude Code", "which extension", "CLAUDE.md vs skill", "subagent vs team", "output style", "custom output", "programmatic Claude", "Agent SDK", "claude -p", "headless mode", "structured output", "CI/CD integration", "MCP vs skill", "hook vs skill", "plugin packaging".

Comprehensive guide to Claude Code hooks — the system for running shell commands, LLM prompts, or subagents automatically at specific points in Claude Code's lifecycle. Use when creating, configuring, debugging, or understanding Claude Code hooks. Trigger phrases: "create a hook", "hook configuration", "hooks.json", "PreToolUse", "PostToolUse", "SessionStart", "hook events", "hook matchers", "automate with hooks", "prompt hooks", "agent hooks", "hook lifecycle", "UserPromptSubmit", "PermissionRequest", "PostToolUseFailure", "SubagentStart", "SubagentStop", "Stop hook", "TeammateIdle", "TaskCompleted", "PreCompact", "SessionEnd", "Notification hook". Also for blocking dangerous commands, auto-formatting code, desktop notifications, file protection, context re-injection after compaction, async hooks, MCP tool hooks, and hook debugging.

Comprehensive guide for setting up and managing MCP (Model Context Protocol) servers in Claude Code. Use when configuring MCP servers, connecting external services via MCP, setting up .mcp.json files, choosing between stdio/HTTP/SSE transports, authenticating with remote MCP servers, bundling MCP servers in plugins, managing MCP scopes (local/project/user), using MCP resources and prompts, configuring managed MCP for organizations, or troubleshooting MCP connections. Trigger phrases: "MCP server", "Model Context Protocol", ".mcp.json", "connect external service", "MCP in plugin", "stdio transport", "SSE transport", "HTTP transport", "remote MCP", "MCP authentication", "MCP tools", "claude mcp add", "MCP OAuth", "MCP scope", "managed MCP", "MCP tool search", "MCP resources", "MCP prompts".

1 hook across 1 event