AI Coding Tool Setup and Reference (2/7) — Claude Code Slash Commands: 83 Commands

From the 5 you use daily to the ones worth knowing about, straight from the official docs


ํ•ต์‹ฌ ์š”์•ฝ

  • Audience: You've used Claude Code but only know a handful of commands beyond /help.
  • What you'll get: All 83 built-in slash commands grouped into 10 categories, the "daily 5" and "worth-knowing 10", and how to author your own custom commands.
  • Prerequisite: Claude Code installed (install guide). Inside a session, typing / auto-filters the full list, but this page serves as the dictionary when you need to map intent to command.

1. The daily 5

These five cover 90% of session time. Learn these first.

Command What it does When to use
/help Show available commands When you forget. Typing / + letters already filters live.
/clear New conversation, empty context When topic fully changes. Old conversation stays in /resume. Aliases: /reset, /new.
/compact [instructions] Summarize conversation to free context Context nearly full but you want to keep working. You can pass focus instructions.
/resume [session] Resume prior conversation After closing the terminal and coming back. Alias: /continue.
/model [model] Switch model Sonnet ↔ Opus. No arg = picker.

/clear vs /compact: /clear starts a fresh conversation (no memory of prior). /compact summarizes and continues (keeps context).


2. Complete list by category

Per the official /commands page, 83 commands across 10 categories. In a session, / filters live; this table is the map when you know what you want but not the name.

2.1 Session control (9)

Command Purpose
/help Help
/clear New conversation (aliases /reset, /new)
/compact [instructions] Summarize and continue
/resume [session] Resume prior conversation (alias /continue)
/branch [name] Branch conversation at this point (alias /fork)
/rewind Rewind to prior point (aliases /checkpoint, /undo)
/context Visualize context usage
/recap One-line session summary
/exit Quit (alias /quit)

2.2 Account & usage (8)

Command Purpose
/login Sign in to Anthropic account
/logout Sign out
/config Settings UI (alias /settings)
/status Config + connectivity
/usage Cost, limits, activity stats (aliases /cost, /stats)
/upgrade Open plan upgrade page
/privacy-settings Privacy settings (Pro/Max)
/extra-usage Configure extra usage past limits

2.3 Model & effort (3)

Command Purpose
/model [model] Switch model. No arg = picker.
/effort [level\|auto] Effort level (low/medium/high/xhigh/max). No arg = slider.
/fast [on\|off] Toggle fast mode

2.4 Project config (8)

Command Purpose
/init Create CLAUDE.md (env CLAUDE_CODE_NEW_INIT=1 for interactive flow)
/add-dir <path> Add a working directory to the session
/memory Edit CLAUDE.md, manage auto-memory
/agents Manage subagent configs
/hooks View hook configs
/mcp MCP servers + OAuth
/skills List skills (press t to sort by token count)
/permissions allow/ask/deny rules (alias /allowed-tools)

2.5 Work tools (8)

Command Purpose
/diff Viewer for uncommitted changes + per-turn diffs
/copy [N] Copy last (or Nth-prior) response. w writes to file
/export [filename] Export conversation as text
/rename [name] Rename session
/insights Session analytics report
/focus Show only last prompt + response (fullscreen only)
/tui [default\|fullscreen] Switch terminal UI renderer
/tasks Background tasks (alias /bashes)

2.6 Plan, review, debug (7)

Command Purpose Type
/plan [description] Enter plan mode built-in
/review [PR] Local PR review built-in
/ultraplan <prompt> Cloud-based deep plan built-in
/ultrareview [PR] Cloud multi-agent review built-in
/security-review Security scan of current branch built-in
/simplify [focus] Review+fix changed files (3 parallel review agents) Skill
/debug [description] Enable debug logging + analysis Skill

2.7 Automation & batch (6)

Command Purpose Type
/batch <instruction> Decompose a large change into 5–30 units, run in parallel Skill
/loop [interval] [prompt] Repeat prompt (alias /proactive) Skill
/schedule [description] Cron routines (alias /routines) built-in
/autofix-pr [prompt] Web session watches branch PR, auto-fixes built-in
/fewer-permission-prompts Scan logs, add allowlist to reduce prompts Skill
/claude-api Load Claude API/SDK reference Skill

Commands marked Skill are bundled skills. Claude reads the skill prompt and orchestrates the work. Claude may also invoke them automatically.

2.8 Plugins, extensions, IDE (10)

Command Purpose
/plugin Manage plugins
/reload-plugins Reload active plugins
/chrome Claude in Chrome settings
/ide IDE integration status
/desktop Switch session to Desktop app (alias /app)
/mobile Mobile app QR (aliases /ios, /android)
/install-github-app Claude GitHub Actions app install
/install-slack-app Slack app install
/remote-control Enable remote control (alias /rc)
/remote-env Default remote env for --remote web sessions

2.9 Theme, environment, keybindings (7)

Command Purpose
/theme Color theme (auto/light/dark/daltonized/ANSI/custom)
/color [color] Prompt bar color
/statusline Status bar config
/terminal-setup Terminal keybindings (VS Code, Cursor, etc.)
/keybindings Open/create keybindings file
/sandbox Toggle sandbox mode
/voice [hold\|tap\|off] Voice dictation toggle

2.10 Misc & utilities (17)

Command Purpose
/bug / /feedback [report] Submit feedback
/heapdump JS heap snapshot to ~/Desktop
/doctor Installation diagnosis (f to auto-fix)
/release-notes Version changelog picker
/team-onboarding Generate team onboarding guide (30d analysis)
/stickers Order stickers
/passes Share 1 free week with friends (if eligible)
/powerup Feature tutorial
/setup-bedrock Bedrock setup (CLAUDE_CODE_USE_BEDROCK=1)
/setup-vertex Vertex setup (CLAUDE_CODE_USE_VERTEX=1)
/web-setup Link GitHub for Claude Code on the web
/teleport Pull web session into this terminal (alias /tp)
/btw <question> Quick side question (not added to conversation)

3. Recently added commands (April 2026)

Per the official release notes, these landed in the last two weeks. If you were away for a few weeks, this list is your upgrade summary.

Version Date New command / feature
v2.1.110 2026-04-15 /tui [default\|fullscreen] — switch terminal UI renderer. fullscreen = flicker-free alt-screen.
v2.1.110 2026-04-15 /focus — focus view showing only last prompt + response (fullscreen only).
v2.1.111 2026-04-16 /effort — no-arg form opens an interactive slider; left/right to pick lowmax, Enter to apply.
v2.1.111 2026-04-16 /ultrareview [PR] — cloud multi-agent deep review. Pro/Max get 3 free runs through 2026-05-05.
v2.1.111 2026-04-16 /fewer-permission-prompts (Skill) — scans transcripts, adds frequent read-only tools to project allowlist automatically.
v2.1.116 2026-04-20 /terminal-setup — only shows in terminals needing it (VS Code/Cursor/Windsurf/Alacritty/Zed). Configures Shift+Enter and friends.
v2.1.118 2026-04-23 /theme — expanded custom theme support. Themes saved to ~/.claude/themes/. Pick "New custom theme…" in the picker.

Removed commands (/pr-comments v2.1.91, /vim v2.1.92) are in section 7.


4. The worth-knowing 10

Past the daily 5, these 10 noticeably raise productivity. Know the names; use when needed.

  1. /context — Color grid of context usage. Spot where tokens leak at a glance.
  2. /rewind — Checkpoint-based undo. Stronger than a single undo.
  3. /branch [name] — Branch the conversation. Useful for A/B approaches.
  4. /copy [N] — Copy Nth-previous response only. Pull a specific section from a long session.
  5. /export [filename] — Export the whole conversation as markdown. Good handoff-doc starter.
  6. /insights — Session analytics. See where your friction actually is.
  7. /recap — One-line summary. For when you need to explain the session to someone else.
  8. /hooks — Current hook configuration. Starting point for hook debugging.
  9. /mcp — MCP servers + OAuth state. First stop when MCP misbehaves.
  10. /usage — Cost, limits, activity. Essential for monthly budget tracking. Same command as /cost//stats.

5. Authoring your own slash commands

If a prompt or procedure repeats, turn it into a custom skill callable as /yourname. Per official docs, .claude/commands/deploy.md and .claude/skills/deploy/SKILL.md both create /deploy (same behavior). Skills are recommended — they support extra features (supporting files, frontmatter control).

5.1 Minimum viable skill

mkdir -p ~/.claude/skills/explain-code

~/.claude/skills/explain-code/SKILL.md:

---
name: explain-code
description: Explains code with visual diagrams and analogies.
---

When explaining code, always include:
1. Start with an everyday analogy
2. Add an ASCII diagram
3. Walk through the code step by step
4. Highlight one common mistake

Then in a session: /explain-code src/auth/login.ts.

5.2 Four places skills can live

Scope Path Applies to
Enterprise managed settings Entire organization
Personal ~/.claude/skills/<name>/SKILL.md All your projects
Project .claude/skills/<name>/SKILL.md This project only
Plugin <plugin>/skills/<name>/SKILL.md Where plugin is enabled

Name conflict priority: enterprise > personal > project.

5.3 Useful frontmatter fields

Field Use
name Slash command name (lowercase, digits, hyphens, max 64)
description Text Claude uses to decide auto-invocation. Capped at 1,536 chars.
disable-model-invocation: true Prevent Claude auto-invoking. Good for side-effect commands like /deploy.
user-invocable: false Hide from user menu (background knowledge only).
allowed-tools Tools the skill can use without per-call approval while active.
arguments, $ARGUMENTS Argument substitution. /fix-issue 123$ARGUMENTS = 123.
context: fork Run in a forked subagent for isolation.

Full list: Skills frontmatter reference.


6. Slash commands vs CLI flags

Often confused, but they operate at different layers.

Aspect Slash command CLI flag
Where to run Inside a session, /... In the terminal, claude ...
Purpose Control session behavior Start sessions / one-off queries
Example /model sonnet, /clear claude --model sonnet, claude -c -p "query"

Many pairs do the same thing (/model--model, /resume--resume, parts of /config). Use CLI flags in scripts/CI, slash commands during live sessions.

Full flag list: official CLI reference.


7. Removed commands (for reference)

You may see these in older blogs or tutorials — they're gone now.

Command Removed Replacement
/pr-comments v2.1.91 Just ask Claude in natural language: "show PR comments"
/vim v2.1.92 /config → Editor mode
--enable-auto-mode v2.1.111 --permission-mode auto

Always check version before trusting an older tutorial.


8. Counter-scenarios — when this classification doesn't fit

  • Some commands don't appear for you → Expected. /desktop is macOS/Windows only; /upgrade is Pro/Max only; /setup-bedrock appears only when the env var is set.
  • /xxxx reports "not a valid command" → Typo first. Claude Code suggests the closest match (/clearr/clear). If not, plugins/skills may have failed to load; /reload-plugins often fixes it.
  • Custom skill doesn't auto-trigger → Check that description contains keywords users would actually type. Inspect the loaded list with /skills. If descriptions are getting truncated, raise the budget with SLASH_COMMAND_TOOL_CHAR_BUDGET.

9. What's next

Once slash commands feel natural:

  1. Claude Code install complete guide — if you're not installed yet or moving to another platform.
  2. CLAUDE.md authoring — what to include, what to avoid — how to fill the file /init created (coming soon).
  3. Claude Code hooks complete guide — 4 lifecycle events × real scripts — move past /hooks viewing into authoring (coming soon).

References


This is post 2/15 in the "AI Coding CLI Entry Guide" series. last verified: 2026-04-24 (per official Claude Code Commands docs).

Series overview: Series index

๋Œ“๊ธ€

์ด ๋ธ”๋กœ๊ทธ์˜ ์ธ๊ธฐ ๊ฒŒ์‹œ๋ฌผ

Agent Memory Engine (2/10) — Building an AI Agent Memory System with SQLite Alone

"ML Foundations (9/9) — PyTorch vs TensorFlow, and the Road to Local LLMs"

"RAG Core Study (14/26) — Evaluation Sets with RAGAS & DeepEval"

"ML Foundations (8/9) — Deep Learning Architectures: CNN, RNN, Attention"

"ML Foundations (7/9) — Deep Learning Training: Optimizers, Regularization, Initialization"

OpenClaw to Hermes Migration (2/13) — What to Preserve, Partially Port, or Discard

AI Agents I Built (5/7) — Building an Automated Blogger API Publishing System