可信 — 风险评分 5/100
上次扫描:19 小时前 重新扫描
5 /100
server-audit
Server Audit Skill for AI Agents — full diagnostic of Proxmox VE and Linux servers (Debian/Ubuntu/RHEL/CentOS) via read-only SSH commands
Pure documentation-only skill describing read-only SSH diagnostic commands for Proxmox and Linux server auditing; no executable code, scripts, or hidden functionality present.
技能名称server-audit
分析耗时41.3s
引擎pi
可以安装
Approve for use. This skill contains only markdown documentation files describing SSH-based diagnostic commands. The skill correctly scopes itself to read-only audit operations with explicit prohibitions against state-changing commands.
资源类型声明权限推断权限状态证据
命令执行 READ READ ✓ 一致 SKILL.md and references/* describe only diagnostic SSH exec commands (lscpu, sma…
文件系统 WRITE WRITE ✓ 一致 linux-audit.md:252 'DOC_DIR=/DATA/local_database/${PROJECT}/Servers/${HOSTNAME}'…
网络访问 READ READ ✓ 一致 All network access is via SSH to remote servers for diagnostic commands only (ip…
环境变量 NONE NONE No environment variable access detected; HOSTNAME/SERVER_IP are captured from th…
技能调用 NONE NONE No skill chaining or inter-skill invocation present
剪贴板 NONE NONE No clipboard access found
浏览器 NONE NONE No browser automation found
数据库 NONE NONE No database access found

目录结构

4 文件 · 29.7 KB · 697 行
Markdown 4f · 697L
├─ 📁 references
│ ├─ 📝 linux-audit.md Markdown 252L · 9.8 KB
│ └─ 📝 proxmox-audit.md Markdown 236L · 9.2 KB
├─ 📝 README.md Markdown 30L · 1.6 KB
└─ 📝 SKILL.md Markdown 179L · 9.1 KB

安全亮点

✓ Skill explicitly scopes to read-only operations ('Аудит = тільки читати') with enumerated prohibitions against state-changing commands
✓ All SSH commands are fully declared in SKILL.md and reference files — no hidden functionality
✓ Commands use standard diagnostic tools (lscpu, dmidecode, smartctl, journalctl, ipmitool, etc.) that match the documented purpose
✓ Documentation saving writes to a predictable local Obsidian vault path — not an external server
✓ No external network connections to unknown IPs; all remote access is over SSH to user-provided server addresses
✓ No obfuscation techniques (base64, eval, atob) present
✓ No credential harvesting, no data exfiltration, no reverse shell indicators
✓ No malicious dependencies — entire file tree is markdown documentation only
✓ Skill is version-controlled and documented with clear trigger conditions and audit workflow