子智能体编排指南：shanraisshan/claude-code-best-practice: claude-sub，适用于工程实践、代码质量与开发流程优化。
子智能体编排指南：shanraisshan/claude-code-best-practice: claude-sub，适用于工程实践、代码质量与开发流程优化。
> 来源: shanraisshan/claude-code-best-practice (64k stars) — Claude Code 完整实践指南
> 原文件: best-practice/claude-subagents.md
> 模型推荐: 看 skill 类型挑
shanraisshan 在 64k stars 的仓库里整理的"Claude Code 完整实践指南" — 涵盖启动参数、commands、MCP、memory、settings、skills、subagents 等所有方面。
michael 强调"skill 要有相应的指导功能,指导用户使用",所以加了下面两节让 Agent 和用户对接。
---
1. 用户问"Claude Code 怎么用 X"时,触发对应 skill
2. skill 严格按原文指南跑
3. 涉及配置改动前 ask user 确认(避免破坏现有配置)
4. 完工后让 Agent 跑自检
1. 问 Agent "Claude Code 怎么 X"(X = 子主题)
2. 如果 Agent 改配置,确认后再继续
3. 全程 Agent 自动化,你只需回答授权类问题
---
 <br>
[](../implementation/claude-subagents-implementation.md)
Claude Code subagents — frontmatter fields and official built-in agent types.
<table width="100%">
<tr>
<td><a href="../">← Back to Claude Code Best Practice</a></td>
<td align="right"><img src="../!/claude-jumping.svg" alt="Claude" width="60" /></td>
</tr>
</table>
---
| Field | Type | Required | Description |
|-------|------|----------|-------------|
| `name` | string | Yes | Unique identifier using lowercase letters and hyphens |
| `description` | string | Yes | When to invoke. Use `"PROACTIVELY"` for auto-invocation by Claude |
| `tools` | string/list | No | Comma-separated allowlist of tools (e.g., `Read, Write, Edit, Bash`). Inherits all tools if omitted. Supports `Agent(agent_type)` syntax to restrict spawnable subagents; the older `Task(agent_type)` alias still works |
| `disallowedTools` | string/list | No | Tools to deny, removed from inherited or specified list |
| `model` | string | No | Model to use: `sonnet`, `opus`, `haiku`, a full model ID (e.g., `claude-opus-4-6`), or `inherit` (default: `inherit`) |
| `permissionMode` | string | No | Permission mode: `default`, `acceptEdits`, `auto`, `dontAsk`, `bypassPermissions`, or `plan` |
| `maxTurns` | integer | No | Maximum number of agentic turns before the subagent stops |
| `skills` | list | No | Skill names to preload into agent context at startup (full content injected, not just made available) |
| `mcpServers` | list | No | MCP servers for this subagent — server name strings or inline `{name: config}` objects |
| `hooks` | object | No | Lifecycle hooks scoped to this subagent. All hook events are supported; `PreToolUse`, `PostToolUse`, and `Stop` are the most common |
| `memory` | string | No | Persistent memory scope: `user`, `project`, or `local` |
| `background` | boolean | No | Set to `true` to always run as a background task (default: `false`) |
| `effort` | string | No | Effort level override when this subagent is active: `low`, `medium`, `high`, `xhigh`, `max` (Opus 4.6 only). Default: inherits from session |
| `isolation` | string | No | Set to `"worktree"` to run in a temporary git worktree (auto-cleaned if no changes) |
| `initialPrompt` | string | No | Auto-submitted as the first user turn when this agent runs as the main session agent (via `--agent` or the `agent` setting). Commands and skills are processed. Prepended to any user-provided prompt |
| `color` | string | No | Display color for the subagent in the task list and transcript: `red`, `blue`, `green`, `yellow`, `purple`, `orange`, `pink`, or `cyan` |
---
| # | Agent | Model | Tools | Description |
|---|-------|-------|-------|-------------|
| 1 | `general-purpose` | inherit | All | Complex multi-step tasks — the default agent type for research, code search, and autonomous work |
| 2 | `Explore` | haiku | Read-only (no Write, Edit) | Fast codebase search and exploration — optimized for finding files, searching code, and answering codebase questions |
| 3 | `Plan` | inherit | Read-only (no Write, Edit) | Pre-planning research in plan mode — explores the codebase and designs implementation approaches before writing code |
| 4 | `statusline-setup` | sonnet | Read, Edit | Configures the user's Claude Code status line setting |
| 5 | `claude-code-guide` | haiku | Glob, Grep, Read, WebFetch, WebSearch | Answers questions about Claude Code features, Agent SDK, and Claude API |
---
本 skill 专注于子智能体编排指南，shanraisshan/claude-code-best-practice: claude-subagents。它将相关流程标准化，帮助用户更快拿到可靠结果，减少重复手工操作。
当你需要在子智能体编排指南相关工作中获得稳定、可复用的产出时最适合——无论是单次任务还是纳入日常工作流，都能直接调用。
需要一个具体的项目或任务上下文，最好带有代码仓库或需求文档。
1. 问 Agent "Claude Code 怎么 X"(X = 子主题)
2. 如果 Agent 改配置,确认后再继续
3. 全程 Agent 自动化,你只需回答授权类问题
---
本 skill 专注于子智能体编排指南，shanraisshan/claude-code-best-practice: claude-subagents。它将相关流程标准化，帮助用户更快拿到可靠结果，减少重复手工操作。
当你需要在子智能体编排指南相关工作中获得稳定、可复用的产出时最适合——无论是单次任务还是纳入日常工作流，都能直接调用。
需要一个具体的项目或任务上下文，最好带有代码仓库或需求文档。