system-health-check（系统健康检查） Skills下载安装命令和教程

安装方式

通过对话安装

复制提示词发给支持本地命令或 Skills 的 AI 助手，先确认命令和权限，再让它执行。

请帮我安装这个 Agent Skill：system-health-check（系统健康检查）
来源仓库：https://github.com/chenxizhang/agent-skills
仓库路径：skills/system-health-check
安装命令：
npx skills add https://github.com/chenxizhang/agent-skills --skill system-health-check
安装前请先检查当前环境是否支持对应 CLI，并向我确认将要执行的命令、安装目录、联网范围和文件读写权限；确认后再执行。

命令行安装

复制命令到本机终端执行。该命令会通过 npx skills 从第三方来源获取 Skill；本站只展示命令，不托管安装包，也不自动执行。

skills.shnpx skills

npx skills add https://github.com/chenxizhang/agent-skills --skill system-health-check

简介

System Health Check 并行扫描系统安全风险、性能瓶颈与优化机会点。

覆盖密码强度、端口暴露、资源占用等全方位健康度评估指标体系。
支持按需启用专项检查模式，适应新机器初始化与生产部署前审计需求。
执行过程不跳过任何阶段，确保全面覆盖不留死角的安全隐患排查。
system-health-check 属于研究检索类 Skill，可作为该场景下的辅助能力补充。

SKILL.md

System Health Check

A comprehensive scanner that analyzes your system for security risks, performance issues, and optimization opportunities — using parallel execution for maximum speed.

When to Use

Setting up a new machine
Periodic security audits
Troubleshooting performance issues
Before deploying to production

Options

Users may request a subset:

Full check (default): All three categories
Security only: Security checks only
Performance only: Performance checks only
Optimization only: Optimization suggestions only

Strict Execution Flow

Do NOT use any scripts. Do NOT skip or merge phases. Execute each phase in order.

Phase 1: Environment Detection (MANDATORY — must display results before proceeding)

Detect and explicitly display the following before doing anything else:

Operating System: Run a command to detect the OS and version.

- Windows: [System.Environment]::OSVersion and (Get-CimInstance Win32_OperatingSystem).Caption - macOS: sw_vers - Linux: cat /etc/os-release | head -5

Shell environment: Identify the current shell.

- PowerShell: $PSVersionTable.PSVersion - bash/zsh: echo $SHELL and version

Agent identity: Identify which agent is running this skill (Claude Code, GitHub Copilot CLI, Cursor, etc.) based on the agent's own context/identity.
Privilege level: Check if running as admin/root.

- Windows PowerShell: ([Security.Principal.WindowsPrincipal][Security.Principal.WindowsIdentity]::GetCurrent()).IsInRole([Security.Principal.WindowsBuiltInRole]::Administrator) - macOS/Linux: whoami (check if root) or id -u (0 = root)

Hostname: hostname

Display the detection results clearly, for example:

Environment Detection:
  OS:        Windows 11 Pro (10.0.22631)
  Shell:     PowerShell 7.4
  Agent:     GitHub Copilot CLI
  Privilege: Standard user (not admin)
  Hostname:  DESKTOP-ABC123

CRITICAL: All subsequent phases MUST use ONLY commands for the detected OS and shell. Never include commands from other platforms — not in execution, not in recommendations, not anywhere.

Phase 2: Plan (generate environment-specific execution plan)

Based on Phase 1 results:

Select commands: From the reference tables below, pick ONLY the column matching the detected OS. Ignore all other columns entirely.
Plan parallelism based on the detected agent:

Agent	Parallel Strategy
GitHub Copilot CLI	Use three sub-agents (task tool with agent_type "task" or "general-purpose") — one for Security, one for Performance, one for Optimization.
Claude Code	Use Agent Teams — dispatch three sub-agents in parallel, one per category.
Other agents	Use whatever parallel execution mechanism is available.

Display the plan before executing, e.g.:

Plan:
  Checks: Security + Performance + Optimization
  Strategy: 3 parallel sub-agents (GitHub Copilot CLI)
  Platform: All commands use PowerShell (Windows)

Phase 3: Execute (parallel)

CRITICAL: The three check categories are COMPLETELY INDEPENDENT. Run them ALL in PARALLEL!

Launch three independent workstreams simultaneously. Each workstream uses ONLY the commands selected in Phase 2 for the detected platform.

Within each workstream, individual checks can also be parallelized for even more speed.

Phase 4: Report & Recommendations

Compile Report

After all parallel workstreams complete, compile a unified report:

================================================================================
                         SYSTEM HEALTH CHECK REPORT
================================================================================
Generated: <timestamp>
System:    <OS and version>
Shell:     <shell and version>
Agent:     <agent identity>
Hostname:  <hostname>

[Security Analysis results...]
[Performance Analysis results...]
[Optimization Suggestions...]

Use severity indicators:

[✓] — Check passed / healthy
[!] — Warning / needs attention
[✗] — Critical issue / security risk

Focus on actionable findings — interpret results, don't dump raw output.

Recommendations

CRITICAL: ALL recommendations MUST be specific to the detected environment.

If on Windows: only recommend PowerShell commands, Windows tools, Windows settings
If on macOS: only recommend macOS commands and tools
If on Linux: only recommend Linux commands and tools
NEVER suggest chmod on Windows. NEVER suggest Get-Acl on Linux. NEVER suggest icacls on macOS. NEVER mix platforms.

Command Reference Tables

The agent MUST only use commands from the column matching the detected OS. Ignore other columns.

🔒 Security Analysis

Check	Windows (PowerShell)	macOS (bash/zsh)	Linux (bash)
Firewall	`Get-NetFirewallProfile`	`defaults read /Library/Preferences/com.apple.alf globalstate`	`ufw status` or `iptables -L`
Open ports	`Get-NetTCPConnection -State Listen`	`lsof -i -P -n \	grep LISTEN`	`ss -tlnp`
SSH config	`Get-Content $env:ProgramData\ssh\sshd_config -ErrorAction SilentlyContinue`	`cat /etc/ssh/sshd_config 2>/dev/null`	`cat /etc/ssh/sshd_config 2>/dev/null`
Updates	`Get-HotFix \	Sort InstalledOn -Desc \	Select -First 5`	`softwareupdate -l`	`apt list --upgradable 2>/dev/null` or `yum check-update`
AV status	`Get-MpComputerStatus`	(skip — not standard)	(skip or check ClamAV)
Users	`Get-LocalUser \	Where-Object Enabled`	`dscl. -list /Users \	grep -v '^_'`	`awk -F: '$3>=1000{print $1}' /etc/passwd`
File permissions	`Get-Acl $env:USERPROFILE\.ssh\* -ErrorAction SilentlyContinue`	`ls -la ~/.ssh/`	`ls -la ~/.ssh/`

AI Agent Security Scanning

Scan these config directories for risky patterns (use the agent's file search capabilities or platform-appropriate grep):

User-level: ~/.claude/, ~/.copilot/, ~/.continue/, ~/.cursor/, ~/.aider/, ~/.agents/, ~/.codeium/, ~/.codeflow/

Project-level: .claude/, .continue/, .cursor/, .copilot/, .github/copilot/

Category	Risk Level	Patterns
Network outbound	HIGH	`curl.POST`, `wget --post`, `Invoke-WebRequest.POST`
Credential access	HIGH	`.ssh/`, `.aws/`, `API_KEY`, `SECRET`, `TOKEN`, `PASSWORD`
Obfuscation	HIGH	`base64 -d`, `base64 --decode`, `String.fromCharCode`
Dynamic execution	MEDIUM	`eval(`, `exec(`, `source <(`, `Function(`
Package installation	MEDIUM	`npx -y`, `pip install` from URLs
Permission bypass	MEDIUM	`bypassPermissions`, `skipVerify`, `dangerouslyAllow`

📊 Performance Analysis

Check	Windows (PowerShell)	macOS (bash/zsh)	Linux (bash)
CPU	`Get-CimInstance Win32_Processor \	Select LoadPercentage `and` Get-Process \	Sort CPU -Desc \	Select -First 5`	`top -l 1 -n 0 \	grep "CPU usage" `and` ps aux --sort=-%cpu \	head -6`	`top -bn1 \	head -5 `and` ps aux --sort=-%cpu \	head -6`
Memory	`Get-CimInstance Win32_OperatingSystem \	Select TotalVisibleMemorySize,FreePhysicalMemory`	`vm_stat` and `sysctl hw.memsize`	`free -h`
Disk	`Get-PSDrive -PSProvider FileSystem \	Select Name,Used,Free`	`df -h`	`df -h`
Network	`(Get-NetTCPConnection).Count`	`netstat -an \	grep ESTABLISHED \	wc -l`	`ss -s`
Processes	`(Get-Process).Count`	`ps aux \	wc -l`	`ps aux \	wc -l`
Startup	`Get-CimInstance Win32_StartupCommand \	Select Name,Command`	`launchctl list \	wc -l`	`systemctl list-unit-files --state=enabled --no-pager`

🔧 Optimization Suggestions

Check	Windows (PowerShell)	macOS (bash/zsh)	Linux (bash)
Temp files	`Get-ChildItem $env:TEMP -Recurse -ErrorAction SilentlyContinue \	Measure-Object -Property Length -Sum`	`du -sh /tmp/ ~/Library/Caches/ 2>/dev/null`	`du -sh /tmp/ /var/tmp/ 2>/dev/null`
Services	`Get-Service \	Where-Object {$_.Status -eq 'Running'} \	Measure-Object`	`launchctl list \	wc -l`	`systemctl list-units --type=service --state=running --no-pager \	wc -l`
Top CPU	`Get-Process \	Sort CPU -Desc \	Select -First 10 Name,CPU,WorkingSet64`	`ps aux --sort=-%cpu \	head -11`	`ps aux --sort=-%cpu \	head -11`
Top Memory	`Get-Process \	Sort WorkingSet64 -Desc \	Select -First 10 Name,@{N='MemMB';E={[math]::Round($_.WorkingSet64/1MB)}}`	`ps aux --sort=-%mem \	head -11`	`ps aux --sort=-%mem \	head -11`

Notes

Some checks may require elevated privileges — if a command fails due to permissions, note it in the report and continue
Adapt commands based on what's actually installed (e.g., if ufw is not available, try iptables; if neither, skip and note)
The command reference tables are hints for the detected platform — the agent may use its knowledge to pick even better available commands, as long as they match the detected OS

适合场景

01

用户想查找某类 Agent Skill 时

02

需要根据任务场景推荐可安装能力包时

03

需要对比不同来源的安装命令和来源信息时

能力概览

能力 1

按任务关键词查找相关 Skills

能力 2

展示可复制的安装命令

能力 3

保留来源站点、仓库和原始说明，方便继续核验

能力 4

展示第三方安全扫描或审计结果

安装后应在对应宿主中按原始 README 的触发条件使用；具体调用方式请以来源页面和 README 为准。

平台分布

Codex

35.41%

按下载量换算42

Claude

29.06%

按下载量换算34

Cursor

20.21%

按下载量换算24

Gemini CLI

9.56%

按下载量换算11

安全审计

Gen Agent Trust Hub

通过

Socket

通过

Snyk

可疑

权限和风险

敏感数据

该 Skill 可能接触密钥、Token、环境变量或敏感配置，应进入高风险复核队列，默认不自动发布。

安装前确认

本站仅展示第三方公开信息，不托管安装包，不提供自动安装或运行环境。安装前应自行审查源码、依赖和命令行为。来源安全扫描存在 warning/failed 结果，不能写成本站确认安全。当前只有一个来源，正式发布前建议补源仓库或其他目录站核验。

来源信息

GitHub 仓库

github.com

/chenxizhang/agent-skills

SKILL.md 来源

github.com

/chenxizhang/agent-skills/tree/main/skills/system-health-check

skills-sh

skills.sh

/chenxizhang/agent-skills/system-health-check

继续浏览同类 Skills

更多研究检索 Skills（73,675）返回 Skills 目录

上一个 Skill

system-health-check（系统健康检查）

system-health-check 用于查找、检索和筛选相关信息，适合在 Codex、Claude、Cursor、Gemini CLI 中需要根据关键词、任务场景或来源线索快速定位候选结果时使用。可结合来源仓库、安装命令和原始 README 继续核验具体用法。安装前建议确认权限范围、维护状态，以及是否会触发联网、命令执行或文件读写。

下一个 Skill

system-integrity-and-backup（系统完整性和备份）

system-integrity-and-backup 用于查找、检索和筛选相关信息，适合在 OpenClaw 中需要根据关键词、任务场景或来源线索快速定位候选结果时使用。可结合来源仓库、安装命令和原始 README 继续核验具体用法。安装前建议确认权限范围、维护状态，以及是否会触发联网、命令执行或文件读写。