The Agent Skills Directory

[COMMAND_EXECUTION]: The Python diagnostic tools check_ceph_health.py and check_cluster_health.py execute remote commands via SSH. The implementation follows security best practices by using subprocess.run with argument lists (disabling shell interpolation), utilizing the -- separator to prevent SSH option injection, and strictly validating node hostnames or IP addresses through regular expressions before execution.
[CREDENTIALS_UNSAFE]: Documentation in api-reference.md and README.md correctly guides users to manage Proxmox API tokens and passwords using environment variables such as PROXMOX_VE_API_TOKEN and PROXMOX_VE_PASSWORD. This approach avoids hardcoding secrets and is consistent with administrative automation standards.
[PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface because it reads and processes data generated by the Proxmox cluster, such as VM names, node descriptions, and CEPH health logs. A maliciously named resource could potentially introduce unexpected instructions into the agent's context during status monitoring. (Ingestion points: cluster_status.py, check_ceph_health.py; Capability inventory: VM and storage management; Sanitization: Regex validation on node names only; Boundary markers: None identified).

proxmox-infrastructure