The Agent Skills Directory

[COMMAND_EXECUTION]: The skill enables remote command execution through SSH-based tools such as ssh, sshpass, paramiko, and fabric. It frequently uses the StrictHostKeyChecking=no option to bypass host key verification, which is a common but notable security trade-off in automated environments to facilitate initial connections.
[EXTERNAL_DOWNLOADS]: The skill provides instructions for installing administrative tools (tmux, sshpass, paramiko, fabric) through official system and language package managers (apt, pip). These are standard utilities used for the skill's primary administrative functions.
[PRIVILEGE_ESCALATION]: The skill covers operations requiring elevated permissions, including the use of sudo for tool installation and the execution of Docker containers in privileged mode.
[INDIRECT_PROMPT_INJECTION]: The skill establishes an attack surface by ingesting data from remote systems (e.g., via tmux capture-pane or reading process output). While this creates a path for a malicious remote server to provide instructions to the agent, it is a functional requirement for the skill's role as a remote management guide.
[CREDENTIALS_UNSAFE]: The skill documents password-based authentication methods, notably using sshpass -p. However, it explicitly mitigates this risk by recommending more secure alternatives such as SSH keys and a 'tmux interactive' method where the user inputs the password directly into their own terminal, keeping it hidden from the AI agent.

remote-server-guide