Diagnose OS-level issues on Linux servers — memory, CPU, disk, network, kernel tuning, resource limits, DNS, time sync, filesystem integrity, browser dependencies, and locale configuration.

This skill helps AI agents systematically diagnose system-level root causes of performance problems, resource exhaustion, and misconfigurations. It is framework-agnostic and contains zero application-specific logic.

Platform: Linux only. This skill is not applicable on Windows or macOS.

What It Checks

System Environment Baseline — OS, kernel, architecture, uptime, load
Memory & OOM — Available memory, total RAM, OOM killer events
CPU & Performance — Load average, CPU utilization, IO wait
Network Infrastructure — DNS resolution, IPv6 configuration, firewall rules
Disk & inotify — Filesystem usage, inotify watch/instance limits
File Descriptor & Process Limits — ulimit, kernel file-nr, per-process nofile
Kernel & Sysctl Tuning — nf_conntrack, somaxconn, TCP tuning, overcommit
DNS Resolution Health — Nameserver configuration, DNS reachability, resolution tests
Time Synchronization — NTP service status, clock drift, hardware clock
Zombie & D-State Processes — Zombie process leaks, uninterruptible sleep, process count
Systemd Journal & Log Disk Usage — Journal size, /var/log growth, large log files
Filesystem Integrity — Read-only mounts, filesystem errors, inode exhaustion
Firewall & Outbound Connectivity — iptables/nftables DROP rules, outbound HTTPS test, ufw
Transparent Hugepages — THP enabled/defrag status, latency impact
TCP Connection Overload — Connection state distribution, CLOSE_WAIT leaks, ephemeral ports
Headless Browser / Chromium Dependencies — Shared library availability, ldd validation, user namespace sandbox
Locale & Encoding Configuration — LANG/LC_ALL settings, locale generation, UTF-8 support

Prerequisites

Linux server (Ubuntu, Debian, CentOS, RHEL, Amazon Linux, etc.)
Root or sudo access
Standard coreutils and system tools: free, df, ps, ss, sysctl, iptables, dmesg, journalctl, timedatectl, ldconfig, ldd, locale, mount, nft, ufw

Usage

Install from ClawHub

clawhub install linux-system-health

Manual

Copy the SKILL.md file into your AI agent's skill directory. For OpenClaw:

mkdir -p ~/.openclaw/skills/linux-system-health
cp SKILL.md ~/.openclaw/skills/linux-system-health/

With other AI agent frameworks

Framework	Method
OpenClaw	`clawhub install linux-system-health`
Qoder	Copy to `.qoder/skills/linux-system-health/SKILL.md`
Claude Code	See below
Cursor	Reference in `.cursorrules`
Codex	Reference in `AGENTS.md`

Claude Code

Method 1: One-command install (recommended)

claude install-skill https://raw.githubusercontent.com/ecsgo-helper/openclaw-system-health/main/linux-system-health/SKILL.md

This installs the skill to .claude/skills/ and makes it available as a slash command in Claude Code sessions.

Method 2: Manual install

mkdir -p .claude/skills
cp SKILL.md .claude/skills/linux-system-health.md

Method 3: Reference in CLAUDE.md (always active)

Add to your project's CLAUDE.md:

## Linux Diagnostics

When diagnosing Linux server issues (performance, memory, disk, network, processes),
follow the diagnostic procedures in `.claude/skills/linux-system-health.md`.

Issue Domains (53 issues)

Domain	Count	Examples
`System`	1	EnvironmentBaseline
`Memory`	4	SystemMemoryCritical, SystemMemoryLow, InsufficientTotalMemory, OOMKillerEvent
`CPU`	3	SystemLoadHigh, SystemCPUExhausted, HighIOWait
`Network`	10	IPv6Mismatch, NoDNSNameservers, DNSResolutionFailed, DNSNameserverUnreachable, FirewallDropRulesDetected, UFWDefaultDeny, TcpConnectionCountHigh, CloseWaitAccumulation, EstablishedConnectionsHigh, EphemeralPortExhaustion
`Disk`	7	FilesystemFull, FilesystemHighUsage, InotifyWatchesTooLow, InotifyInstancesTooLow, ReadOnlyFilesystem, InodeUsageHigh, FilesystemErrorsDetected
`Limits`	3	NofileTooLow, NofileExceedsKernelMax, SystemFileDescriptorsHigh
`Kernel`	10	NfConntrackMaxTooLow, NfConntrackTableFull, SomaxconnTooLow, TcpMaxTwBucketsTooLow, TcpTwReuseNotEnabled, TimeWaitOverflow, TcpListenOverflows, StrictOvercommitWithLowSwap, THPEnabled, THPDefragEnabled
`Time`	3	NTPServiceNotRunning, ClockNotSynchronized, ClockDriftDetected
`Process`	3	ZombieProcessesHigh, DStateProcessesFound, TotalProcessCountHigh
`Logs`	2	JournalDiskUsageHigh, VarLogOversized
`Browser`	3	ChromiumDependenciesMissing, ChromiumBinaryLddFailures, UserNamespaceDisabled
`Locale`	3	LocaleNotConfigured, LocaleNotGenerated, NonUTF8LocaleDetected

Severity Levels

Level	Meaning
FATAL	System cannot support workloads at all
CRITICAL	Major functionality broken or at risk
ERROR	Partial service degradation
WARNING	Potential risk under load
INFO	Informational, no action needed

openclaw-diagnostic — OpenClaw-specific diagnostics (installation, gateway, API connectivity, service management, application logs)

License

MIT-0

README.md

linux-system-health