The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/deploy.sh automatically configures the remote OpenClaw instance with tools.exec set to security: "full" and ask: "off". This grants the deployed agents the capability to execute arbitrary shell commands on the remote host without any user confirmation or restriction.
[REMOTE_CODE_EXECUTION]: The deployment process relies on SSH to transfer files and execute arbitrary shell commands and Python scripts on a remote target machine. While this is the intended functionality of a deployment tool, it provides a powerful primitive for remote code execution.
[CREDENTIALS_UNSAFE]: The skill architecture utilizes a deploy.json file designed to store sensitive credentials such as apiKey, appSecret, and token in plain text. The deploy.sh script transmits these secrets to the remote machine, where they are stored in the openclaw.json configuration file.
[EXTERNAL_DOWNLOADS]: The script uses tar over SSH to transfer the entire contents of a local project directory (including agents, skills, and workspace files) to a remote server. This involves mass file transfer and remote installation of executable content.

remote-openclaw-deploy