| name | claude-slash-commands |
| description | Authoritative, version-pinned reference for every built-in Claude Code slash (/) command — what each does, its aliases, and which are bundled Skills/Workflows vs fixed built-ins — plus how custom slash commands work (the commands-merged-into-skills model, file locations, frontmatter, arguments, !bash, @file). Includes the full /loop vs /schedule breakdown. Use when any session needs to know what a slash command does, find the right command for a task (clear context, switch model, review code, schedule work, background a task), or author a custom /command. Trigger: "claude slash commands", "what does /X do", "list the slash commands", "built-in commands", "custom slash command", "/loop vs /schedule", "how do I make a /command", "/claude-slash-commands". Sibling of claude-cli-flags (launch flags). |
claude-slash-commands — built-in / command reference
Pinned to claude 2.1.183 (source: code.claude.com/docs/en/commands, verified against the binary).
A command is only recognized at the start of a message; text after the name is passed as arguments.
Type / in-session to see what's available to you (plan/platform-gated commands vary). Re-derive
the canonical list anytime by fetching the commands reference; re-check if the version moved past 2.1.183.
Skill = prompt-based, Claude orchestrates with tools. Workflow = multi-agent fan-out.
Everything else = fixed built-in logic. Aliases shown as (= /x).
Session & context (args: session)
| Command | Does |
|---|
/clear [name] | New conversation, empty context; old one stays in /resume. (= /reset, /new) |
/compact [instructions] | Summarize the conversation to free context (optional focus) |
/context [all] | Visualize context-window usage as a grid + optimization hints |
/rewind | Rewind conversation and/or code to a prior point. (= /checkpoint, /undo) |
/resume [session] | Resume by id/name or open picker. (= /continue) |
/branch [name] | Fork the conversation at this point to try another direction; original kept |
/fork <directive> | Spawn a forked subagent that inherits the convo; result returns to you (v2.1.161+) |
/btw <question> | Quick side question that doesn't bloat history |
/export [filename] | Export the conversation as plain text |
/copy [N] | Copy the last (or Nth-latest) assistant response to clipboard |
/rename [name] | Rename the current session shown on the prompt bar |
/recap | One-line summary of the current session |
/diff | Interactive diff viewer: git diff and per-turn diffs |
/cd <path> | Move the session to a new working dir, cache preserved (v2.1.169+) |
/add-dir <path> | Grant file access to another dir for this session |
Model & reasoning (args: model)
| Command | Does |
|---|
/model [model] | Switch model + save as default; s on a row = session-only |
/effort [level|auto] | Set reasoning effort (low/medium/high/xhigh/max/auto) |
/fast [on|off] | Toggle fast mode (faster Opus output) |
/plan [description] | Enter plan mode, optionally with a starting task |
/goal [condition|clear] | Run until a condition is met (until-done) |
Project setup, permissions & config (args: project)
| Command | Does |
|---|
/init | Generate a starter CLAUDE.md (set CLAUDE_CODE_NEW_INIT=1 for interactive) |
/memory | Edit CLAUDE.md memory files; manage auto-memory |
/permissions | Manage allow/ask/deny tool rules. (= /allowed-tools) |
/config [key=value …] | Settings UI, or set a key directly (v2.1.181+). (= /settings) |
/hooks | View hook configurations for tool events |
/mcp [reconnect|enable|disable <server>] | Manage MCP servers |
/agents | Manage subagent configurations |
/skills | List skills; hide a skill from Claude or the / menu |
/plugin [subcommand] | Manage plugins (list/install/enable/disable) |
/reload-plugins [--force] | Reload plugins without restart |
/reload-skills | Re-scan skill/command dirs mid-session (v2.1.152+) |
/keybindings | Open the keyboard-shortcuts file |
/statusline | Configure the status line |
/terminal-setup | Configure Shift+Enter etc. (only in terminals that need it) |
/sandbox | Toggle sandbox mode (supported platforms) |
Parallel & background work (args: parallel)
| Command | Does |
|---|
/background [prompt] | Detach the session to run as a background agent. (= /bg) |
/tasks | View/manage everything running in the background. (= /bashes) |
/stop | Stop the attached background session (keeps transcript/worktree) |
/batch <instruction> | Skill. Decompose a big change into 5–30 units, one subagent+worktree each, opens PRs |
/workflows | Open the workflow progress view (watch/pause/resume/save) |
Code review, run & quality (args: review)
| Command | Does |
|---|
/review [PR] | Review a PR locally in this session |
/code-review [low|medium|high|ultra] | Skill. Review the diff for bugs; ultra = deep cloud multi-agent |
/security-review | Analyze branch changes for security vulnerabilities |
/simplify [target] | Skill. Parallel cleanup review + apply fixes (no bug-hunt) (v2.1.154+) |
/run | Skill. Launch & drive your app to see a change work (v2.1.145+) |
/verify | Skill. Build+run the app to confirm a change works (v2.1.145+) |
/run-skill-generator | Skill. Record a per-project run/verify recipe |
/debug [description] | Skill. Enable debug logging + troubleshoot from the log |
/fewer-permission-prompts | Skill. Mine transcripts → add a read-only allowlist |
/deep-research <question> | Workflow. Fan-out web research → cited report |
/ultraplan <prompt> | Draft a plan in a cloud session, review in browser, execute |
Cloud / web / remote (args: cloud)
| Command | Does |
|---|
/remote-control | Make this session drivable from claude.ai. (= /rc) |
/teleport | Pull a Claude-Code-on-web session into this terminal. (= /tp) |
/background | (see Parallel) detaches to a background/cloud agent |
/autofix-pr [prompt] | Cloud session that watches the branch's PR and pushes CI/review fixes |
/desktop | Continue this session in the Claude Code Desktop app. (= /app) |
/remote-env | Choose the default environment for cloud agents |
/web-setup | Connect GitHub to Claude Code on the web via local gh |
/install-github-app | Set up the Claude GitHub Actions app for a repo |
/install-slack-app | Install the Claude Slack app (OAuth) |
Scheduling & automation (args: schedule) — see also the loop-vs-schedule section
| Command | Does |
|---|
/loop [interval] [prompt] | Skill. Re-run a prompt while the session stays open; omit interval → self-paced; omit prompt → .claude/loop.md or maintenance. (= /proactive) |
/schedule [description] | Create/update/list/run cloud Routines (Anthropic-managed infra). (= /routines) |
Account, usage & info (args: account)
| Command | Does |
|---|
/login · /logout | Sign in / out of your Anthropic account |
/usage | Session cost, plan limits, activity (by skill/subagent/plugin/MCP). (= /cost, /stats) |
/usage-credits | Configure credits to keep working past a limit |
/status | Settings → Status tab: version, model, account, connectivity |
/doctor | Diagnose the install/settings; press f to let Claude fix |
/insights · /team-onboarding | Reports analyzing your sessions / onboarding guide |
/release-notes | Changelog version picker |
/help | Help + available commands |
/feedback [report] | Submit feedback / bug / share. (= /bug, /share) |
/privacy-settings | View/update privacy settings (Pro/Max) |
/upgrade · /passes | Plan upgrade / share a free week |
/exit | Exit the CLI (detaches if attached to a bg session). (= /quit) |
UI / misc (args: ui)
/theme · /color · /tui [default\|fullscreen] · /focus · /scroll-speed · /vim (removed v2.1.92 → use /config Editor mode) ·
/voice · /radio (Claude FM) · /powerup · /mobile (/ios,/android) · /stickers · /heapdump · /chrome · /ide ·
/setup-bedrock · /setup-vertex (only with the matching env flag).
/loop vs /schedule — the distinction that matters (args: loop-vs-schedule)
| /loop | /schedule (cloud Routine) |
|---|
| Lifetime | session-scoped; dies on exit; 7-day auto-expire | durable; persists indefinitely |
| Runs where | this live terminal session (only while REPL idle) | Anthropic cloud infra |
| Survives machine-off? | no | yes |
| Local Mac access (files/shell/MCP)? | yes | no (cloud sandbox) |
| Min interval | self-paced 1 min–1 hr (or any fixed interval) | 1 hour |
| Prompt context | inherits the live session | must be fully self-contained |
| Backing tool | CronCreate / ScheduleWakeup | create_scheduled_task |
| Best for | poll / iterate / watch while you're here | automate on a cadence even when you're gone |
Decision rule: must outlive the session or run with the Mac off → /schedule. Needs to keep going
while you're working (and may touch local files) → /loop. Need "the instant X changes" → the Monitor
tool (streams, not polls), not either of these.
Third option — local scheduled task: definitions at ~/.claude/scheduled-tasks/<name>/SKILL.md, run by
the Claude Desktop app (NOT the terminal's /schedule, NOT a launchd job by default). Local Mac access
- ~1-min cadence, but needs the machine on and the Desktop app running. For true always-on local automation,
an OS-level scheduler (e.g. a launchd plist on macOS, or a systemd/cron unit on Linux) is the real mechanism.
Custom slash commands (args: custom)
Custom commands are now Skills. .claude/commands/deploy.md and .claude/skills/deploy/SKILL.md both
create /deploy. Old .claude/commands/ files keep working; skills add a folder + frontmatter + auto-loading.
Where they live → name:
| Location | /name |
|---|
~/.claude/skills/<dir>/SKILL.md (personal) or .claude/skills/<dir>/SKILL.md (project) | dir name |
.claude/commands/<file>.md | file name |
Plugin skills/<dir>/ | plugin-name:dir |
| Nested project skill on name clash | apps/web:deploy |
Precedence: enterprise > personal > project; a same-name skill overrides a bundled command; a skill beats a command of the same name.
Frontmatter (SKILL.md): description (required) · allowed-tools · argument-hint · model (or inherit) ·
disable-model-invocation: true (only you can run it — use for side-effecting /deploy, /commit) ·
user-invocable: false (only Claude can load it — background knowledge).
Arguments & dynamics in the body:
$ARGUMENTS → the full argument string as typed.$1, $2, … (or $ARGUMENTS[0]…) → positional args (shell-style quoting; quote multi-word values).! → run bash and inline its output into the prompt (e.g. pull a live git diff).
@path → inline a file's contents.
Invocation: type /name directly (if user-invocable), or Claude auto-loads it when relevant (unless
disable-model-invocation). A few built-ins (/init, /review, /security-review) are also exposed to Claude via the Skill tool; most (e.g. /compact) are not.
Honesty
- List pinned to v2.1.183; visibility varies by plan/platform (e.g.
/privacy-settings Pro/Max; /setup-bedrock only with CLAUDE_CODE_USE_BEDROCK=1). Type / in-session for your actual set.
/vim and /pr-comments were removed (v2.1.92 / v2.1.91); listed for recognition only.- Descriptions condensed from the official reference — open it for the full prose and version notes.
