A Claude Code plugin for debugging Linux desktops — targeted journal/boot/log inspection skills, plus an idempotent installer that instruments the system with the proactive logging tools needed to catch hard crashes and analyze them after the fact.
Targets Ubuntu + Wayland desktops. Most skills work on any systemd Linux; the install script assumes apt. Forkable for other distros.
Run scripts/install-debugging-stack.sh once to set up:
| Tool | Purpose |
|---|---|
| Persistent journald | journalctl -b -1 survives reboot — logs available after a hard crash |
linux-crashdump (kdump-tools) |
Full kernel core dumps on panic via kexec, written to /var/crash/ |
sysstat |
Records CPU / memory / I/O / network every 10 min — sar lets you look back at system state before a freeze |
systemd-coredump |
Captures userspace segfault core dumps |
pstore check |
Verifies firmware-backed crash storage as a belt-and-braces fallback |
Userspace OOM protection (systemd-oomd / earlyoom) is handled separately by the setup-oom-protection skill.
journal-inspect— Targetedjournalctlqueries by unit, time, priority, kernel scope, with structured/JSON output for AI analysis. The default replacement forjournalctl -xe.boot-inspect— Failed units,systemd-analyze blame/critical-chain, kernel boot messages, boot-to-boot diff. Wayland session services included.log-inspect— Non-journal log sources: app state dirs (~/.cache,~/.local/state),/var/log/*, dpkg/apt history, snap/flatpak. Uses ripgrep + jq for AI-friendly output.
setup-oom-protection— Configure systemd-oomd (Ubuntu 22.04+) or earlyoom so the desktop kills runaway processes before the kernel hard-locks under memory pressure.inspect-oom-events— Find past OOM kills (kernel + oomd + earlyoom) in the journal and reconstruct what was running, correlated with sar memory/swap data.
validate-persistent-journal— Verify/var/log/journalis active so logs survive reboot. Configure if not. Precondition for any post-crash investigation.investigate-last-crash— Systematic post-mortem after a hard crash: enumerate boots, identify the crash window, collect kernel + userspace evidence, correlate with sar metrics, write a findings report.setup-netconsole— Stream kernel oops messages over UDP to another LAN host, for capturing freezes that leave no local log trail.review-pstore— Inspect/sys/fs/pstorefor firmware-saved panic logs that survive reboots.
From the Claude Code marketplace:
claude plugins install linux-debugging@danielrosehillThen, optionally, run the proactive instrumentation installer:
git clone https://github.com/danielrosehill/Linux-Debugging-Plugin
cd Linux-Debugging-Plugin
bash scripts/install-debugging-stack.shA reboot is required for kdump to become active.
Invoke skills naturally in a Claude Code session:
"Inspect the journal for errors in the last hour" "Why is boot taking 45 seconds?" "Set up OOM protection on this machine" "Investigate the last crash"
Ubuntu (tested on Ubuntu 25.10 with KDE Plasma on Wayland). Most skills are systemd-portable; the install script is apt-only. Fork and adapt the installer for other distros.
MIT