How to use the Debian Linux Triage agent skill

What the Debian Linux Triage skill makes the agent do

An agent skill is a folder with a SKILL.md file that tells the agent how to approach a certain kind of task. This skill covers Debian Linux issues using apt, systemd and AppArmor-aware guidance. Once it is loaded, the agent first confirms your Debian release and setup, then lays out a numbered triage plan using systemctl, journalctl, apt and dpkg.

For fixes, the agent gives copy-paste-ready commands, adds verification commands after each major change, notes AppArmor or firewall considerations where relevant, and provides rollback or cleanup steps. The skill only shapes the agent's guidance; it contains no code of its own.

• Summary of the problem
• Numbered triage steps
• Remediation commands in code blocks
• Validation commands after changes
• Rollback or cleanup steps

Install the skill from the ZIP file

Download the ZIP for skills/debian-linux-triage from this page and unzip it. The package contains a single file, SKILL.md. Put the unzipped folder into one of two places. For a personal skill that applies to all your projects, use ~/.claude/skills/. For a skill that only applies to one project, use the .claude/skills/ folder inside that project.

If you use the Claude apps instead, you can upload the skill under Settings → Capabilities → Skills. Choose whichever route matches how you work; the skill's content is the same either way.

• Personal install: ~/.claude/skills/
• Project install: .claude/skills/ inside your project
• Claude apps: Settings → Capabilities → Skills

Trigger the skill with natural requests

You do not need a special command. Describe your Debian problem in plain words and include the details the agent will need. The source notes that the skill takes an optional Debian release, a problem summary and optional constraints, so mention the release if you know it and any rules you must follow, such as not restarting a service during business hours.

If the agent asks follow-up questions about your release or environment, answer them. The skill tells the agent to ask concise questions when it needs them before giving a plan.

• Mention your Debian release if you know it (for example, the output of your system's release information).
• State the symptom: what fails, what you expected, and any error text.
• Name constraints such as 'do not remove packages' or 'this is a test machine'.

Examples of good requests

Good requests name the service or package, the symptom and the environment. The agent can then pick the right systemctl and journalctl checks rather than guessing.

• 'My nginx service on Debian fails to start after an upgrade. Help me triage it with systemctl and journalctl and tell me how to roll back.'
• 'apt install stopped with a dpkg error on my Debian server. Give me a numbered plan and verification steps before changing anything.'
• 'My custom systemd service starts but exits right away. Read the logs and suggest a fix, and note if AppArmor could be blocking it.'
• 'I am new to Debian. Explain each command you suggest and how to check it worked.'

Safety: review commands before you run them

The skill has no scripts to run and does not use the network, so it does not run code on your machine by itself. The commands it suggests are run by you, or by the agent in your session, so the safety question is about those commands. Fixes such as package removal or service changes can cause outages or data loss if applied blindly.

Treat the output as AI-generated advice that must be checked. Test on a non-production machine first, read each command before running it, keep backups, and use the rollback steps the agent provides. For critical infrastructure, have an experienced administrator review the plan.

• Not designed for Fedora, Arch or other non-Debian systems, where package names and tools differ.
• Do not run remediation commands unreviewed on production systems.
• Check the rollback steps before you start a change.

Tips for better results

Give the agent the raw evidence. Paste the relevant journalctl output or the exact apt error rather than a paraphrase. Ask for the verification command after each change so you can confirm one thing at a time. If a fix does not work, tell the agent what happened and ask for the next step instead of stacking several changes together.

• Paste exact error messages and log lines.
• Change one thing at a time and run the validation commands between changes.
• Ask for an undo path before any removal or service change.