The Agent Skills Directory

[COMMAND_EXECUTION]: The skill makes extensive use of sudo to execute administrative commands such as virsh, lxc, and snap. These are required for managing hypervisors and system containers on the host.
[EXTERNAL_DOWNLOADS]: Documentation workflows suggest the use of rclone to transfer container backups to external cloud storage providers (e.g., Google Drive). This is described as an optional manual step for off-host recovery.
[INDIRECT_PROMPT_INJECTION]: The skill possesses an attack surface for indirect prompt injection by ingesting untrusted data from the managed environments.
Ingestion points: Container logs and status information are read via lxc info --show-log and lxc info (SKILL.md, references/lxd-reference.md).
Boundary markers: None identified in the provided instructions.
Capability inventory: The skill can execute arbitrary commands inside containers (lxc exec), modify configurations (lxc config set), and perform destructive actions like container deletion (lxc delete --force) or VM destruction (virsh destroy).
Sanitization: No specific sanitization or filtering of container-generated output is mentioned before being processed by the agent.

linux-virtualization