The Agent Skills Directory

[COMMAND_EXECUTION]: In TOOL.py, the execute_tool function uses subprocess.run with shell=True on a string formatted with user input. This allows an attacker to execute arbitrary shell commands by injecting metacharacters into the command parameter.
[COMMAND_EXECUTION]: In TOOL.py, the process_file function uses os.system to run a cat command with an unvalidated filename. This is vulnerable to command injection through the filename variable.
[REMOTE_CODE_EXECUTION]: In TOOL.py, the run_user_script function uses the exec() function to run code read from a file path provided as an argument. This enables arbitrary Python code execution if the file path points to a malicious script.

command_injection_test