← Back to Workshop
>_ Claude Code Dojo
Master every slash command, shortcut, and flag. Study the reference or train in the dojo.
Open PDF Reference
Score: 0
Slash Commands
/helpShow help and available commands
/clear /reset /newClear conversation and free context
/compact [focus]Compact conversation, optional focus area
/model [name]Switch model (opus, sonnet, haiku)
/fast [on|off]Toggle fast output mode
/planEnter plan mode (read-only exploration)
/initInitialize project with CLAUDE.md
/memoryEdit CLAUDE.md and auto-memory files
/config /settingsOpen settings interface
/permissionsView/update tool permissions
/diffOpen interactive diff viewer
/commitStage and commit changes
/reviewReview a pull request
/pr-comments [PR]Fetch GitHub PR comments
/costShow token usage statistics
/usageShow plan limits and rate status
/statusShow version, model, account info
/contextVisualize context usage as grid
/copyCopy last response to clipboard
/export [file]Export conversation as text
/resume /continueResume a previous session
/fork [name]Create fork of conversation
/rewind /checkpointRewind to a previous point
/add-dir <path>Add a working directory
/mcpManage MCP server connections
/hooksManage hook configurations
/keybindingsOpen keybindings config
/vimToggle vim editing mode
/themeChange color theme
/terminal-setupConfigure terminal keybindings
/doctorDiagnose installation issues
/sandboxToggle sandbox mode
/tasksList background tasks
/statsDaily usage visualization & streaks
/insightsGenerate session analysis report
/exit /quitExit Claude Code
Bundled Skills
/simplifyAuto-review and fix recent code changes
/batch <instruction>Large-scale parallel changes across codebase
/debug [description]Troubleshoot Claude Code session issues
/loop [interval] <prompt>Run a prompt repeatedly on a schedule
/claude-apiLoad Claude API reference material
Keyboard Shortcuts
General Controls
Ctrl+CCancel current input or generation
Ctrl+DExit Claude Code (EOF)
Ctrl+GOpen prompt in external editor
Ctrl+LClear screen (keeps history)
Ctrl+OToggle verbose output (see tool usage)
Ctrl+RReverse search command history
Ctrl+BBackground running tasks
Ctrl+TToggle task list visibility
Ctrl+FKill background agents (press twice)
Ctrl+V / Cmd+VPaste image from clipboard
Text Editing
Ctrl+KDelete to end of line
Ctrl+UDelete entire line
Ctrl+YPaste deleted text (from Ctrl+K/U)
Alt+B / Alt+FMove cursor back/forward one word
Mode & Model Switching
Shift+Tab / Alt+MCycle permission modes (Normal > Auto > Plan)
Alt+P / Option+PSwitch model without clearing prompt
Alt+T / Option+TToggle extended thinking
Esc + EscRewind / restore to previous point
Multiline Input
\ + EnterNewline (works in all terminals)
Option+EnterNewline (macOS native)
Shift+EnterNewline (iTerm2, WezTerm, Ghostty, Kitty)
Ctrl+JLine feed
CLI Flags & Options
Session Control
claudeStart interactive session
claude "query"Start with initial prompt
claude -p "query"Print response and exit (headless)
claude -cResume most recent conversation
claude -r "name"Resume specific session by name/ID
claude -w feature-nameStart in isolated git worktree
Model & Behavior
--model <name>Set model: opus, sonnet, haiku, or full ID
--fast [on|off]Toggle fast mode
--permission-mode <mode>plan, auto-accept, or normal
--max-turns <n>Limit agentic turns (print mode)
--max-budget-usd <n>Spend limit before stopping (print mode)
--verboseShow detailed turn-by-turn output
Context & Input
--add-dir <path...>Add extra working directories
--append-system-prompt <text>Add text to system prompt
--system-prompt <text>Replace entire system prompt
--json-schema <schema>Get validated JSON output (print mode)
Output Formats
--output-format textPlain text output (default)
--output-format jsonJSON output with metadata
--output-format stream-jsonStreaming JSON events
Permissions & Safety
--dangerously-skip-permissionsSkip all permission prompts
--allowedTools "Bash(git *)"Auto-allow specific tools
--disallowedTools "Bash(curl *)"Block specific tools
MCP & Integration
--mcp-config <path>Load MCP servers from file
--chrome / --no-chromeEnable/disable Chrome automation
--agent <name>Use a specific agent type
Special Syntax
! <command>Bash mode - execute shell command directly
@file.jsMention file (includes content in context)
@src/Mention directory (includes listing)
@server:protocol://pathReference MCP resource
/ at start of inputCommand/skill autocomplete
TabAccept prompt suggestion
"ultrathink" in promptTrigger high-effort reasoning turn
MCP Subcommands
claude mcp listList all configured MCP servers
claude mcp get <name>Get server details
claude mcp remove <name>Remove a server
claude mcp add --transport http <n> <url>Add HTTP server
claude mcp add --transport stdio <n> -- <cmd>Add stdio server
claude mcp add-json <name> '<json>'Add server from JSON
claude mcp add-from-claude-desktopImport from desktop app
claude mcp serveUse Claude as an MCP server
Authentication
claude auth loginSign in to Anthropic account
claude auth login --ssoSign in via SSO
claude auth logoutSign out
claude auth statusCheck auth status (JSON)
claude auth status --textCheck auth status (readable)
Environment Variables
API & Model
ANTHROPIC_API_KEYAPI key for direct access
ANTHROPIC_MODELDefault model (e.g., claude-opus-4-6)
CLAUDE_CODE_EFFORT_LEVELReasoning depth: low | medium | high
MAX_THINKING_TOKENSFixed thinking budget limit
Behavior
CLAUDE_CODE_MAX_OUTPUT_TOKENSMax output tokens (up to 64000)
CLAUDE_CODE_SHELLOverride shell detection
CLAUDE_CODE_DISABLE_FAST_MODEDisable fast mode (set to 1)
CLAUDE_CODE_DISABLE_GIT_INSTRUCTIONSRemove git context (set to 1)
DISABLE_AUTOUPDATERDisable auto-updates (set to 1)
Cloud Providers
CLAUDE_CODE_USE_BEDROCK=trueUse AWS Bedrock backend
CLAUDE_CODE_USE_VERTEX=trueUse Google Vertex AI backend
Configuration Files
~/.claude/settings.jsonUser settings (all projects)
.claude/settings.jsonProject settings (shared/committed)
.claude/settings.local.jsonProject settings (gitignored)
~/.claude/keybindings.jsonCustom keyboard shortcuts
CLAUDE.md / .claude/CLAUDE.mdProject instructions for Claude
.claude/rules/Rule files (path-specific context)
.mcp.jsonProject MCP servers (shared)
.claude/hooks.jsonProject hooks configuration
Permission Modes
NormalAsks permission for each tool use
Auto-AcceptAuto-approves configured tools
PlanRead-only exploration, asks before changes
Shift+Tab to cycleNormal → Auto-Accept → Plan → Normal
Hook Events
PreToolUseBefore any tool executes
PostToolUseAfter a tool succeeds
PostToolUseFailureAfter a tool fails
UserPromptSubmitBefore processing user prompt
SessionStartWhen session initializes
SessionEndWhen session ends
NotificationWhen Claude needs attention
StopWhen session is stopping
SubagentStart / SubagentStopSubagent lifecycle events
TaskCompletedBackground task finished
PreCompactBefore context compaction
WorktreeCreate / WorktreeRemoveWorktree lifecycle
ConfigChangeSettings changed
Tips & Hidden Features
! npm testRun shell commands directly (bash mode)
/compact focus on authCompact with specific focus area
/doctorDeep diagnostic of your installation
/statsSee your usage streaks & patterns
/insightsAI-generated session analysis
claude mcp serveUse Claude itself as an MCP server
claude --init-onlyRun init hooks then exit immediately
Tab on suggestionAccept grayed-out prompt suggestion
Ctrl+R then typeSearch through prompt history
CLAUDE_CODE_TASK_LIST_ID=xNamed persistent task list