When working with Claude Code (Claude CLI), certain flags are powerful but verbose. Typing them repeatedly slows you down and increases the chance of mistakes.

A common example is running Claude with relaxed permission checks:

claude --dangerously-skip-permissions

This guide shows how to create custom commands specifically for Claude Code, so you can run Claude in different modes with short, readable commands.

---

🎯 Goal

Turn this:

claude --dangerously-skip-permissions prompt.txt

Into this:

claude-dev prompt.txt

Same behavior. Cleaner workflow.

---

🧠 How Claude Code Is Typically Used

Claude Code is often run in different “modes”, such as:

• Development mode
• Experimental mode
• Safe/default mode
• Debug or verbose mode

Each mode usually means the same claude command with different flags. This makes Claude an ideal candidate for custom wrapper commands.

---

✅ Why Use Wrapper Commands Instead of Aliases

While aliases work, wrapper scripts are the better choice for Claude Code because:

• They behave like real commands
• They accept arguments cleanly
• They work in scripts, SSH sessions, and automation
• They make Claude usage self-documenting

For Claude Code, this is the recommended pattern.

---

🛠 Creating a Claude-Specific Command Wrapper

Step 1: Decide the Claude mode name

Choose a name that clearly represents how Claude will run:

claude-dev
claude-exp
claude-unsafe
claude-fast

In this example, we’ll use:

claude-dev

---

Step 2: Create the command

Run:

sudo tee /usr/local/bin/claude-dev > /dev/null <<'EOF'
#!/bin/bash
claude --dangerously-skip-permissions "$@"
EOF

What this does:

• Wraps the claude binary
• Always applies the chosen flags
• Passes all user arguments through correctly

---

Step 3: Make it executable

sudo chmod +x /usr/local/bin/claude-dev

---

Step 4: Test it

claude-dev

Or with a file:

claude-dev prompt.txt

Claude now runs in your custom “dev” mode automatically.

---

⚠️ Important: Correct Flag Name

Claude CLI is strict about flag names.

❌ Incorrect:

--dangerous-skip-permissions

✅ Correct:

--dangerously-skip-permissions

If you’re unsure, always verify with:

claude --help

---

🔍 Verifying Your Claude Wrapper

Confirm which command is running:

which claude-dev

Expected output:

/usr/local/bin/claude-dev

Inspect the wrapper:

cat /usr/local/bin/claude-dev

---

💡 Recommended Claude Command Patterns

Once you understand the pattern, you can create multiple Claude-specific commands:

Safe default Claude

claude-safe

#!/bin/bash
claude "$@"

---

Experimental Claude

claude-exp

#!/bin/bash
claude --dangerously-skip-permissions --verbose "$@"

---

Project-specific Claude

claude-project-x

#!/bin/bash
claude --context project-x "$@"

Each command becomes a clear mental model for how Claude will behave.

---

🧠 Why This Pattern Works Well for Claude Code

Claude Code is:

• Flag-heavy
• Context-sensitive
• Often run repeatedly with the same options

Wrapper commands turn Claude into a toolbox of modes, instead of a single overloaded command.

This scales extremely well as your usage grows.

---

🏁 Conclusion

If you use Claude Code seriously, creating Claude-specific wrapper commands is one of the best quality-of-life improvements you can make.

You get:

• Shorter commands
• Fewer mistakes
• Clear intent
• Better automation

And most importantly — a smoother Claude workflow.