Claude Code 基本用法与哲学基本用法 Quick Reference Keyboard Shortcuts S

基本用法

Quick Reference

Keyboard Shortcuts

Shortcut	Action
`!command`	Execute bash immediately
`Esc Esc`	Rewind conversation/code
`Ctrl+R`	Reverse search history
`Ctrl+S`	Stash current prompt
`Shift+Tab`(×2)	Toggle plan mode
`Alt+P`/ `Option+P`	Switch model
`Ctrl+O`	Toggle verbose mode
`Tab`/ `Enter`	Accept prompt suggestion

Essential Commands

Command	Purpose
`/init`	Generate CLAUDE.md for your project
`/context`	View token consumption
`/stats`	View your usage statistics
`/usage`	Check rate limits
`/vim`	Enable vim mode
`/config`	Open configuration
`/hooks`	Configure lifecycle hooks
`/sandbox`	Set permission boundaries
`/export`	Export conversation to markdown
`/resume`	Resume a past session
`/rename`	Name current session
`/theme`	Open theme picker
`/terminal-setup`	Configure terminal integration

CLI Flags

Flag	Purpose
`-p "prompt"`	Headless/print mode
`--continue`	Resume last session
`--resume`	Pick a session to resume
`--resume name`	Resume session by name
`--teleport id`	Resume a web session
`--dangerously-skip-permissions`	YOLO mode

SKILLS

quick-feature

/quick-feature 迭代分支名需求描述

analyze-component

专门用来分析1000行以上大组件的

core pattern

                    THE AGENT PATTERN
                    =================

    User --> messages[] --> LLM --> response
                                      |
                            stop_reason == "tool_use"?
                           /                          \
                         yes                           no
                          |                             |
                    execute tools                    return text
                    append results
                    loop back -----------------> messages[]


    That's the minimal loop. Every AI coding agent needs this loop.
    Production agents add policy, permissions, and lifecycle layers.

用户输入 → 消息数组 → LLM 处理 → 生成响应
                           ↓
              stop_reason == "tool_use"?
             /                          \
           是                           否
            |                             |
        执行工具                     返回文本结果
        将结果追加到消息
        循环回去 ───────────────→ 消息数组

def agent_loop(messages):
    while True:                                    # ① 持续循环直到不需要工具
        response = client.messages.create(         # ② 调用 Claude API
            model=MODEL,                           #    如: "claude-sonnet-4-6"
            system=SYSTEM,                         #    系统提示词
            messages=messages,                     #    完整对话历史
            tools=TOOLS,                           #    可用工具定义
        )
        
        # ③ 将 LLM 的响应添加到历史
        messages.append({
            "role": "assistant",
            "content": response.content
        })

        # ④ 检查是否需要调用工具
        if response.stop_reason != "tool_use":
            return  # 如果不需要工具，直接返回结果
        
        # ⑤ 执行所有工具调用
        results = []
        for block in response.content:
            if block.type == "tool_use":
                output = TOOL_HANDLERS[block.name](**block.input)
                results.append({
                    "type": "tool_result",
                    "tool_use_id": block.id,
                    "content": output,
                })
        
        # ⑥ 将工具结果追加到历史，循环继续
        messages.append({
            "role": "user",
            "content": results
        })