Keyboard / UI
shift-tab # normal- / auto-accept- / planning-Mode
ctrl-R # detailed output
esc-esc # jump to previous message
ctrl-esc # quick launch from ide (not working!)
/commands (built-in)
/clear # Clear conversation history
/ide
/cost # Show costs of current session
/agents # Manage agents
@filename.txt # @file mention
# always write unit tests # Add to memory
!ls # Run bash commands
/init # initialize CLAUDE.md
/output-style # Default, Explanatory, Learning
/commands (custom)
/command1 [argument] # .claude/commands/command1.md (project command)
/command2 [argument] # ~/.claude/commands/command2.md (user command)
# commands in subfolders
/frontend:test # .claude/commands/frontend/test.md
# arguments
/fix-issue 123 # $ARGUMENTS <- 123
/commands (mcp)
/mcp # list mcp servers
/mcp__github__list_prs
Memory file (CLAUDE.md)
IMPORTANT # emphasis
@README.md # link to files
thinking
think
think more, think a lot, think harder, think longer
ultrathink
CLI
claude --dangerously-skip-permissions
claude [options] [command] [prompt]
claude -p # print response and exit
claude -p "Whats my age again?"
Mechanisms
Claude Code Features Comparison
| Feature |
Invocation |
Context |
Use When
|
| Sub-Agents |
Automatic OR explicit |
Separate context window per agent |
You need specialized AI for specific task types with focused tools/prompts
|
| Slash Commands |
Explicit (`/command`) |
Adds to current conversation |
You repeat the same prompt frequently and want quick access
|
| Skills |
Automatic (Claude decides) |
Progressive disclosure (loads files as needed) |
Complex multi-file capability that Claude should discover automatically
|
| CLAUDE.md |
Automatic (loaded at startup) |
Always in context |
Persistent memory, preferences, conventions across sessions
|
| Output Styles |
Manual (`/output-style`) |
Replaces system prompt |
Change Claude's overall personality/behavior for entire session
|