The Agent Skills Directory

[COMMAND_EXECUTION]: The tools tools/check_cluster_health.py and tools/check_ceph_health.py utilize subprocess.run to execute shell commands (pvecm, ceph, corosync-cfgtool) on remote Proxmox nodes via SSH. This is the primary method used to gather diagnostics not available via the standard API.
[PRIVILEGE_ESCALATION]: The skill requires administrative access to the Proxmox environment. Documentation and workflows (workflows/cluster-formation.md, workflows/ceph-deployment.md) involve the use of become: true in Ansible and root access via SSH to perform sensitive tasks like cluster initialization, disk wiping (wipefs), and logical volume management.
[DYNAMIC_EXECUTION]: Diagnostic scripts dynamically construct shell commands to be executed over SSH. For instance, check_ceph_health.py interpolates pool names and OSD IDs into command strings. While the scripts include basic regex validation for node addresses, the dynamic construction of commands represents a standard management pattern with inherent risks if input data is compromised.
[INDIRECT_PROMPT_INJECTION]: The skill ingests data from the Proxmox API and remote shell outputs (e.g., VM names, storage pool names, and task statuses). This creates an attack surface where a compromised infrastructure or maliciously named resource could attempt to influence the agent's logic through the data it processes.

proxmox-infrastructure