The Agent Skills Directory

[COMMAND_EXECUTION]: The skill is designed to execute shell commands and code on remote Matrix OS instances using the matrix CLI and MCP tools like run_command and create_terminal. This is the core functionality and is handled via structured commands.
[INDIRECT_PROMPT_INJECTION]: Because the skill reads external data from chat history (get_chat) and remote files (read_file), it is theoretically susceptible to indirect prompt injection. However, the skill incorporates several defense-in-depth measures: it instructs the agent to normalize all paths and reject .. segments (preventing directory traversal) and requires passing user-influenced strings as separate arguments in an argv array to avoid shell command injection vulnerabilities.

matrix-cloud-run