Safety Levels

Magic Shell includes a comprehensive safety analysis system that categorizes commands by risk level and can block dangerous operations.

Severity Levels

Every command is analyzed and assigned a severity level:

Low

Commands worth reviewing but generally safe.

git checkout main
npm install
ls -la

Behavior: Executed without confirmation in all safety modes.

Medium

Commands requiring elevated privileges or making notable changes.

sudo apt install nginx
chmod 755 script.sh
rm -rf node_modules/

Behavior: May require confirmation depending on safety level.

High

Commands with significant risk of data loss or system changes.

sudo rm -rf /var/log/*
kill -9 -1
shutdown -h now

Behavior: Requires confirmation in moderate and strict modes.

Critical

Commands that could cause irreversible system damage. These are always blocked.

rm -rf /           # Delete entire filesystem
:(){ :|:& };:      # Fork bomb
dd if=/dev/zero of=/dev/sda  # Overwrite disk
chmod -R 777 /     # Recursive permission change on root

Safety Modes

Configure your preferred safety level:

Strict

Confirm all potentially risky commands (low and above).

msh --safety strict

Best for:

Learning environments
Production servers
Shared systems

Moderate (Default)

Confirm high severity commands; critical commands are always blocked.

msh --safety moderate

Best for:

Daily development work
Personal machines
Balanced protection

Relaxed

Skip confirmations for most commands; critical commands are still blocked.

msh --safety relaxed

Best for:

Experienced users
Quick scripting
Trusted environments

Blocked Patterns

Certain dangerous patterns are always blocked, regardless of safety level:

Pattern	Risk
`rm -rf /`	Delete entire filesystem
`rm -rf ~`	Delete home directory
`:(){ :\|:& };:`	Fork bomb
`> /dev/sda`	Overwrite disk
`mkfs.*`	Format filesystem
`wget \| sh`	Pipe download to shell
`curl \| bash`	Pipe download to shell
`chmod -R 777 /`	Recursive root permission change

How Analysis Works

Magic Shell analyzes commands using:

Pattern Matching: Regex patterns detect known dangerous commands
Privilege Detection: Commands using sudo, doas, etc. are flagged
Path Analysis: Operations on sensitive paths (/, /etc, /usr) are scrutinized
Combination Detection: Dangerous flag combinations are identified

Dry-Run Mode

Use dry run to see safety analysis without executing:

msh -n "delete all log files older than 30 days"

Output includes:

The translated command
Severity level
Matched patterns (if any)
Whether confirmation would be required

Customization

Blocked Commands

Add custom patterns to block in your config:

{
  "blockedCommands": [
    "DROP TABLE",
    "TRUNCATE TABLE"
  ]
}

Confirmed Patterns

Track patterns you’ve previously confirmed:

{
  "confirmedDangerousPatterns": [
    "rm -rf node_modules"
  ]
}

Best Practices

Start with strict mode when learning or on new systems
Use dry run (-n) for unfamiliar queries
Review commands before using -x flag
Keep moderate mode for daily work
Never bypass safety for commands you don’t understand