Shell Command Tool

⚠️ IMPORTANT RESTRICTION: You CANNOT use this tool directly! To execute shell commands, you MUST first create a subagent (using create_subagent skill) and then run it (using run_subagent skill). Direct execution of shell commands is not allowed.

execute_shell_command: A bridge to the local operating system shell. Allows execution of standard Bash commands.

Usage

from tools import execute_shell_command

# List files
output = execute_shell_command("ls -la")

# Read a file
content = execute_shell_command("cat -n README.md")

Parameters

Returns

Returns a single string containing the command's standard output (stdout) and standard error (stderr) combined.

Note: 1. Safety Interception: Commands like rm, rmdir, shred are strictly forbidden and will return an error string. 2. Output Truncation: If the output exceeds 5000 characters, it will be truncated. Use head, tail, or grep for large files.

Example

from tools import execute_shell_command

# 1. Inspect current directory
print(execute_shell_command("ls -F"))

# 2. Install a package (if network is available)
print(execute_shell_command("pip install pandas"))

# 3. Read a specific part of a file (safe for large files)
print(execute_shell_command("head -n 20 data.csv"))

Tips

• Non-Interactive: Do NOT use interactive tools like vim, nano, or python (without -c). The process will hang.
• Reading Files: Prefer cat -n filename to see line numbers, which is helpful for editing later.
• Large Files: Avoid cat on large files. Use head -n 50 filename to peek, or grep to search.
• Environment: You are in a standard Bash environment. You can use pipes: ls -la | grep .py.
• State: The working directory is persistent. cd folder works for subsequent commands.
• Environment Setup: Use this tool to install Python packages and other dependencies needed by your subagent. For example: pip install playwright && playwright install chromium, pip install pandas, etc.