CLSkills
April 8, 2026Claude Skills Hub

Complete List of Claude Code Slash Commands (2026 Edition)

Every Claude Code slash command in one place — built-in commands, hidden ones, and the secret prompt codes that work like commands. Updated for 2026 with examples for each.

claude-codeslash-commandsclaudecheat-sheet2026AI
📬

Get notified when we discover new Claude codes

We test new prompt commands every week. Join 4+ developers getting them in their inbox.

Why Slash Commands Matter

Claude Code is a CLI. Slash commands are how you steer it. Most users only know /help and /clear, but there are dozens of built-in commands and hundreds of community-discovered prompt codes that behave like commands.

This is the complete reference, organized by what each one actually does.

Built-In Claude Code Commands

These ship with Claude Code by default. Type any of them at the start of your message.

Session Management

  • /clear — Clears the current conversation context. Use it when Claude starts mixing up files from earlier in the session.
  • /compact — Compresses earlier messages into a summary. Use it when you're approaching context limits but want to keep going.
  • /exit — Quits Claude Code cleanly.
  • /restart — Restarts the current session, preserves loaded skills and CLAUDE.md.
  • /cost — Shows token usage and cost for the current session.

File and Project

  • /init — Generates a CLAUDE.md file for your current project. Run this once per project.
  • /status — Shows what files Claude has loaded and what tools are active.
  • /files — Lists files Claude has access to in the current session.
  • /diff — Shows the diff of any uncommitted changes Claude has made.
  • /undo — Reverts the last file edit Claude made.

Workflow

  • /plan — Forces Claude into Plan Mode — it explains its approach before executing anything. Critical for non-trivial tasks.
  • /approve — Approves the plan from /plan and starts execution.
  • /skip — Skips the current step and moves to the next one.
  • /retry — Retries the last action that failed.
  • /explain — Asks Claude to explain its last response in more detail.

Tools

  • /tools — Lists every tool Claude can call (Bash, Read, Write, Edit, etc).
  • /web — Forces Claude to search the web before answering.
  • /git — Stages git operations (commit, push, branch).
  • /test — Runs tests after Claude makes changes.

Help

  • /help — Shows the built-in command list.
  • /docs — Opens the Claude Code documentation.
  • /version — Shows the installed Claude Code version.

Skill Commands

When you put a markdown file in ~/.claude/skills/, Claude Code loads it as persistent context. You can then invoke skills like commands:

  • /skill <skill-name> — Activates a specific skill from your skills folder.
  • /skills — Lists all skills currently loaded.
  • /skill reload — Reloads skills without restarting the session.

Browse 2,300+ free skills to see what's possible.

Community-Discovered Prompt Codes

These aren't built into Claude Code, but Claude consistently recognizes them and changes its response when you use them at the start of a prompt. Some are user-discovered conventions, some feel almost like easter eggs.

Reasoning Modes

  • L99 — Pushes Claude to maximum reasoning depth. Useful for complex architectural decisions. Full L99 explainer →
  • /deepthink — Forces longer reasoning chain before answering.
  • OODA — Observe, orient, decide, act framework. Great for decisions.
  • INVERT — Solves problems by thinking backward from the goal.
  • XRAY — Sees through the obvious answer to find the real one.

Output Modes

  • /ghost — Removes all meta-commentary and self-references. Pure output.
  • /raw — Strips formatting, gives clean plain text.
  • ARTIFACTS — Structures output as a numbered list of deliverables.
  • /godmode — Most comprehensive, detailed response possible.
  • BEASTMODE — Pushes output quality to maximum.

Style Modes

  • /mirror — Matches your writing style from a sample.
  • /punch — Makes every sentence hit harder, more direct.
  • /polish — Cleans up rough drafts.
  • /trim — Removes fluff without losing meaning.

Coding Modes

  • /debug — Finds bugs and explains them clearly.
  • REFACTOR — Cleans up messy code without changing behavior.
  • ARCHITECT — Designs system structure before any code is written.
  • /optimize — Makes code faster without breaking it.
  • /testit — Writes tests for the code in context.

See all 100+ prompt codes →

Custom Commands (Make Your Own)

Claude Code lets you create your own commands by adding files to ~/.claude/commands/. Each file becomes a slash command:

# ~/.claude/commands/ship.md
Review the current diff. Run tests. If tests pass, commit with a clear message and push to main.

Now /ship runs that workflow whenever you type it.

This is one of the most underused features in Claude Code. You can build a personal command library that turns repetitive workflows into one-line invocations.

Which Commands Actually Get Used?

From the most active Claude Code users we've talked to, the daily-driver commands are:

  1. /plan — used before every non-trivial task
  2. /clear — to reset context when files get mixed up
  3. /skill <name> — to invoke project-specific expertise
  4. /diff — before approving any commit
  5. L99 — for high-stakes architectural questions

If you only learn five, learn those.

Stay Updated

New prompt codes get discovered every week. We test them and add the ones that actually work to clskills.in/prompts — searchable, filterable, copy-to-clipboard.

If you want them in your inbox as we find them, subscribe to the newsletter at the top or bottom of this page.

One email a week. Zero fluff.

New Claude Code skills, hidden prompt codes, and tested workflows — straight to your inbox. No spam, unsubscribe in 1 click.