The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill fetches and executes a setup script from the official NodeSource domain (deb.nodesource.com) to install Node.js. This is an established and well-known service for distributing Node.js environments.
[EXTERNAL_DOWNLOADS]: Downloads configuration scripts and software packages from trusted or well-known sources, including NodeSource and official Ubuntu repositories, via curl, wget, and apt.
[COMMAND_EXECUTION]: The skill generates and executes several system administration commands, such as pct create, apt upgrade, and systemctl, which are required for container lifecycle management.
[PROMPT_INJECTION]: The skill identifies a surface for indirect prompt injection by embedding untrusted user data directly into shell commands without sanitization.
Ingestion points: User-supplied variables such as {service-name}, {IP}, and {package} in SKILL.md.
Boundary markers: No explicit delimiters or boundary instructions are present to prevent embedded malicious sequences in user input from altering command logic.
Capability inventory: The skill allows for significant system-level operations, including creating containers on the host, modifying systemd services, and writing to protected directories like /opt/ and /etc/.
Sanitization: There is no evidence of input validation, escaping, or filtering of user-provided strings before they are used as command-line arguments.

proxmox-lxc