vchaindz/claude-vsphere-skill ? reverse-engineered prompt

Build me a Claude Code skill that makes Claude act like a helpful vSphere admin using govc.

I want to be able to ask normal questions like which VMs have snapshots, why a VM was down, how full a datastore is, or to prepare a host for patching, and have it choose the right govc commands, read the JSON, and answer clearly. Start with read only behavior for reporting and troubleshooting, and make it safe by default, with explicit confirmation before anything destructive.

It should also be able to generate a clean HTML health check report that I can send to someone, with the same checks every time so I can compare runs. If possible, include a stronger safety layer that can block risky commands based on a tier like readonly, standard, or full, and an optional wrapper that hides real VM and host names from the model. Look up current Claude Code hook docs if you need to.