code-with-claude

Comprehensive Claude Code CLI reference covering interactive mode, CLI flags, commands, environment variables, hooks, channels, checkpointing, plugins, and tools. Use when asked about Claude Code features, configuration, workflows, keyboard shortcuts, session management, permission modes, environment setup, or best practices for using Claude Code effectively.

Use this skill when the user asks about Claude Code CLI features, configuration, workflows, or best practices.

Claude Code Feature Map

Feature	What It Does	Reference
CLI flags	Launch-time options for sessions, models, permissions, prompts	./references/cli-reference.md
Commands	In-session / commands for model switching, permissions, context, workflows	./references/commands.md
Interactive mode	Keyboard shortcuts, Vim mode, multiline input, voice, task list	./references/interactive-mode.md
Hooks	Shell commands, HTTP endpoints, or LLM prompts at lifecycle events	./references/hooks.md
Plugins	Extend Claude Code with skills, agents, hooks, MCP/LSP servers	./references/plugins.md
Channels	MCP servers that push external events into a session	./references/channels.md
Checkpointing	Automatic file edit tracking with rewind and summarize	./references/checkpointing.md
Tools	Built-in tools: Bash, Edit, Read, Write, Agent, WebFetch, etc.	./references/tools.md
Environment variables	Control behavior via env vars or settings.json	./references/env-vars.md

Quick Reference: Essential CLI Flags

Flag	Purpose	Example
--model	Set model (alias or full name)	claude --model opus
--continue, -c	Resume most recent conversation	claude -c
--resume, -r	Resume by session ID or name	claude -r "auth-refactor"
--print, -p	Non-interactive print mode (SDK)	claude -p "query"
--permission-mode	Start in a specific permission mode	claude --permission-mode plan
--worktree, -w	Isolated git worktree session	claude -w feature-auth
--bare	Minimal mode, skip auto-discovery	claude --bare -p "query"
--append-system-prompt	Add to default system prompt	claude --append-system-prompt "Use TS"
--effort	Set effort level for session	claude --effort high

Quick Reference: Essential Commands

Command	Purpose
/model [model]	Switch model mid-session
/compact [instructions]	Compact conversation, free context
/clear	Clear conversation history
/rewind	Rewind to a checkpoint (Esc+Esc)
/permissions	Manage tool allow/deny rules
/plan [description]	Enter plan mode
/batch <instruction>	Parallel large-scale changes
/hooks	View hook configurations
/memory	Edit CLAUDE.md memory files
/diff	Interactive diff viewer
/cost	Show token usage statistics
/effort [level]	Set effort (low/medium/high/max)

Quick Reference: Key Shortcuts

Shortcut	Action
Ctrl+C	Cancel current generation
Esc+Esc	Rewind / checkpoint menu
Shift+Tab	Cycle permission modes
Ctrl+O	Toggle transcript viewer
Ctrl+B	Background running tasks
Ctrl+T	Toggle task list
Alt+P	Switch model
Alt+T	Toggle extended thinking
!command	Run bash directly in session
@path	File path autocomplete

Quick Reference: Permission Modes

Mode	Behavior
default	Ask for each tool use
acceptEdits	Auto-approve file edits, ask for bash
plan	Read-only exploration, no writes
auto	AI classifier approves safe operations
bypassPermissions	Skip all permission prompts

Cycle modes with Shift+Tab. Start in a specific mode: claude --permission-mode plan.

Quick Reference: Core Tools

Tool	Description	Permission
Bash	Execute shell commands	Yes
Read	Read file contents	No
Edit	Targeted file edits	Yes
Write	Create or overwrite files	Yes
Agent	Spawn subagent with own context	No
Glob	Find files by pattern	No
Grep	Search file contents	No
WebFetch	Fetch URL content	Yes
WebSearch	Web search	Yes
Monitor	Background watch + react	Yes

Gotchas

• Env vars don't persist across Bash tool commands. Use CLAUDE_ENV_FILE or a SessionStart hook.
• Checkpoints don't track bash changes — only direct file edits via Edit/Write tools are tracked.
• applyTo: "**" burns context — always-on instructions load on every interaction. Use specific globs.
• macOS Alt shortcuts require Option-as-Meta in your terminal (iTerm2: Profiles → Keys → "Esc+").
• --bare disables everything — hooks, skills, plugins, MCP, auto-memory, CLAUDE.md. Use for fast scripted calls.
• Hook if patterns use tool permission syntax: Bash(rm *) matches bash commands starting with rm.
• Plugin agents can't use hooks, mcpServers, or permissionMode frontmatter for security reasons.

When to Load References

Load the appropriate reference when the user needs detailed information:

• CLI flags, system prompt customization, launch options → Read ./references/cli-reference.md
• In-session commands, slash commands, skills list → Read ./references/commands.md
• Keyboard shortcuts, Vim mode, multiline, voice, task list → Read ./references/interactive-mode.md
• Hooks lifecycle, events, matchers, configuration → Read ./references/hooks.md
• Plugins structure, manifest, MCP/LSP, distribution → Read ./references/plugins.md
• Channels, webhook receivers, two-way chat bridges → Read ./references/channels.md
• Checkpointing, rewind, restore, summarize → Read ./references/checkpointing.md
• Built-in tools, Bash behavior, LSP, Monitor, PowerShell → Read ./references/tools.md
• Environment variables, timeouts, model config, proxy → Read ./references/env-vars.md