Slash commands and built-in shortcuts
/model, /effort, /context, /memory, /agents — the built-in commands that change how the session behaves. Plus how to write your own.
Slash commands
Claude Code has ~20 built-in slash commands. Three are worth memorizing:
/model— switch model mid-session (Opus → Sonnet for cheaper edits)/effort— set reasoning budget (low / medium / high / max / xhigh on Opus 4.7+)/context— visualize how much of the window you've used/config— persistent across sessions since Claude Code 2.1.120 (April 2026)
Plus: /memory, /agents, /mcp, /hooks, /skills — list / inspect / edit each subsystem.
Step 1: Try the basics
In any Claude Code session:
/context
You see a colored grid of token usage. Helpful when you're wondering whether to start a fresh session.
/model
Pick one. Sonnet 4.6 for most edits, Opus 4.7 for hard reasoning. The fast / slow tradeoff is real.
Step 2: Write your first custom slash command
Custom commands are markdown files in ~/.claude/commands/<name>.md. The frontmatter sets a description, the body is the prompt that gets injected.
mkdir -p ~/.claude/commands
cat > ~/.claude/commands/standup.md <<'EOF'
---
description: Brief morning standup — what was yesterday, what's today
---
Tell me three things, in order:
1. What I worked on yesterday (last 24h of git activity).
2. What's blocking me right now (open TODOs, failing tests).
3. The single most important task for today.
Keep it under 200 words. No marketing language.
EOF
Now /standup works in any session.
Step 3: Verify
Run aiguide_validate_step — checks ~/.claude/settings.json parses (slash commands don't need settings.json themselves but the validator confirms your environment is sane).
Custom commands beat skills when the trigger is a fixed slash word and you don't want fuzzy matching. Use skills (Recipe 1.2) when you want natural-language triggers.