The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The installation guide in data/install.md provides a method for piped remote execution (curl -fsSL https://get.x-cmd.com | sh). This method is explicitly identified in the documentation as high risk because it executes remote code without manual review.\n- [EXTERNAL_DOWNLOADS]: The tool is designed to automatically download required backends (such as 7zip or zstd) at runtime to ensure operation in minimal environments. Additionally, the installation guide references downloads from https://get.x-cmd.com, https://github.com/x-cmd/release, and https://conda.prefix.dev.\n- [PERSISTENCE]: The installation process modifies shell configuration files (~/.bashrc and ~/.zshrc) to load the x-cmd environment, ensuring the tool remains active across user sessions.\n- [COMMAND_EXECUTION]: The skill's primary function is to execute shell commands (x z, x uz, x zuz ls) for archive processing. The agent is encouraged to use these aliases to manage files and directories.\n- [INDIRECT_PROMPT_INJECTION]: The skill is designed to process external archive files which may contain untrusted data.\n
Ingestion points: The x uz (decompress) and x zuz ls (list) commands process user-provided archive files.\n
Boundary markers: The instructions do not specify any delimiters or warnings to ignore potential instructions embedded in file names or archive metadata.\n
Capability inventory: The tool can execute shell commands and perform file system operations (write/delete).\n
Sanitization: There is no mention of sanitizing or validating the contents or structure of archives before processing.

x-zuz