可信 — 风险评分 0/100
上次扫描:2 天前 重新扫描
0 /100
vmware-monitor
Read-only VMware vCenter/ESXi monitoring skill with 8 MCP tools and CLI for listing VMs, hosts, datastores, clusters, alarms, and events
vmware-monitor is a legitimate read-only VMware vCenter/ESXi monitoring tool. All documented capabilities are read-only; no destructive code paths exist in the codebase. No credential harvesting, exfiltration, or hidden functionality detected.
技能名称vmware-monitor
分析耗时28.5s
引擎pi
可以安装
This skill is safe to use. Follow the documented installation method (uv tool install). Always use a read-only vSphere service account for credentials to minimize blast radius.
资源类型声明权限推断权限状态证据
文件系统 NONE READ ✓ 一致 Reads ~/.vmware-monitor/config.yaml and ~/.vmware-monitor/.env for configuration…
网络访问 NONE READ ✓ 一致 Connects to vCenter/ESXi via pyVmomi SOAP API (documented read-only monitoring)
命令执行 WRITE WRITE ✓ 一致 Bash tool used to invoke vmware-monitor CLI (documented, minimal surface)
环境变量 NONE READ ✓ 一致 Reads VMWARE_MONITOR_CONFIG and .env for target config (documented)
技能调用 NONE NONE No cross-skill invocation observed
剪贴板 NONE NONE No clipboard access
浏览器 NONE NONE No browser access
数据库 NONE READ ✓ 一致 Reads audit.db via vmware-policy for logging (documented, local SQLite only)

目录结构

5 文件 · 22.3 KB · 548 行
Markdown 4f · 510L JSON 1f · 38L
├─ 📁 evals
│ └─ 📋 evals.json JSON 38L · 1.2 KB
├─ 📁 references
│ ├─ 📝 capabilities.md Markdown 80L · 3.3 KB
│ ├─ 📝 cli-reference.md Markdown 93L · 2.6 KB
│ └─ 📝 setup-guide.md Markdown 147L · 6.5 KB
└─ 📝 SKILL.md Markdown 190L · 8.6 KB

安全亮点

✓ Zero destructive code paths — vmware-monitor explicitly does not contain power, create, delete, snapshot, or modify operations
✓ All operations audited to ~/.vmware/audit.db via vmware-policy decorator
✓ Webhook notifications send only to user-configured HTTP endpoints (Slack/Discord) with no credential or PII exfiltration
✓ Source code is fully open source (MIT) on GitHub — reviewable before deployment
✓ Credentials stored in ~/.vmware-monitor/.env (chmod 600) with no storage of passwords in config.yaml
✓ Prompt injection protection: vSphere event/log content is truncated, stripped of control chars, and wrapped in boundary markers
✓ MCP tools are all read-only (list_virtual_machines, list_esxi_hosts, list_all_datastores, list_all_clusters, get_alarms, get_events, vm_info)
✓ CLI uses uv tool install with version pinning from PyPI — no curl|bash or wget|sh remote script execution
✓ TLS verification enabled by default; disableSslCertValidation only for isolated lab environments with self-signed certs
✓ Suggests using a read-only vSphere service account, minimizing blast radius if credentials are compromised