The Agent Skills Directory

[COMMAND_EXECUTION]: The skill documents tools that enable arbitrary command execution on the management host and guest systems.
proxmox_lxc_exec (references/proxmox-lxc.md) allows executing commands on Proxmox nodes via SSH to manage LXC containers.
proxmox_agent_exec (references/proxmox-vm-lxc-shared.md) allows executing shell commands inside QEMU VMs using the guest agent.
[EXTERNAL_DOWNLOADS]: The proxmox_storage_content tool (references/proxmox-storage.md) facilitates downloading files from user-specified external URLs directly to the cluster's storage backends.
[CREDENTIALS_UNSAFE]: The skill instructions (SKILL.md) describe the use of the user's private SSH key located at ~/.ssh/id_rsa for node authentication and container execution.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from multiple sources which could be used for injection attacks.
Ingestion points: Tools such as proxmox_node_log (references/proxmox-nodes.md), proxmox_node_task, and proxmox_agent_exec read logs and process outputs from the managed infrastructure.
Boundary markers: The documentation does not specify the use of delimiters or 'ignore' instructions when processing these outputs.
Capability inventory: The skill possesses broad capabilities including file writes (proxmox_agent_file), guest deletion (proxmox_guest_delete), and network reconfiguration.
Sanitization: There is no documented validation or escaping for data retrieved from guest agents or system logs.

proxmox-mcp-tools