Build a hands-on Linux admin exam to test whether an agent can do your job
Working on an agent to replace himself at work, a Linux admin built his own benchmark in the style of the RHCSA: one jumphost, three servers and 45 tasks, a third of them traps such as prompts hidden in log files.
Source
@BlueTeamWarrior on X · 2026-10-19
Written in our own words. Credit to the original author. Link checked 2026-10-19.
What they did
Working on an agent to replace himself at work, a Linux admin built his own benchmark in the style of the RHCSA: one jumphost, three servers and 45 tasks, a third of them traps such as prompts hidden in log files.
Result (as claimed)
A realistic test of agents on sysadmin work, with a full breakdown.
Tools they used
Cost: Pricing depends on the tools you pick; check each official pricing page.
Set it up
Pick a contained repo or task
Start on a branch or side project where mistakes are cheap.
Install and configure
Set up Linux. Read what it can run on your machine and limit permissions where possible.
Run it on a real task
Give it a well-scoped issue and let it work. Review the diff like any pull request.
Add guardrails
Keep tests and CI in the loop, and never let an agent push to main or touch production secrets unattended.
Copy the full setup prompt
Paste into your agent and it will walk you through the setup, asking before anything that sends, posts or spends.
You are helping me set up this AI use case end to end in Claude Code. Goal: Build a hands-on Linux admin exam to test whether an agent can do your job What the original author did: Working on an agent to replace himself at work, a Linux admin built his own benchmark in the style of the RHCSA: one jumphost, three servers and 45 tasks, a third of them traps such as prompts hidden in log files. Reported result (as claimed by the author): A realistic test of agents on sysadmin work, with a full breakdown. Source: https://x.com/BlueTeamWarrior/status/2112027547764670729 Tools they used: Linux Category: Dev & ops Please: 1. Ask me 3 short questions about my situation (accounts I have, budget, what I want the output to look like). 2. Propose a step-by-step plan using these tools or cheaper alternatives, with the time and cost of each step. Use official pricing pages; say "check pricing" if unsure. 3. Then do the setup with me one step at a time, waiting for my OK before anything that sends messages, posts publicly, spends money or deletes data. 4. Finish with a checklist to run it again next time.
You are helping me set up this AI use case end to end in Codex. Goal: Build a hands-on Linux admin exam to test whether an agent can do your job What the original author did: Working on an agent to replace himself at work, a Linux admin built his own benchmark in the style of the RHCSA: one jumphost, three servers and 45 tasks, a third of them traps such as prompts hidden in log files. Reported result (as claimed by the author): A realistic test of agents on sysadmin work, with a full breakdown. Source: https://x.com/BlueTeamWarrior/status/2112027547764670729 Tools they used: Linux Category: Dev & ops Please: 1. Ask me 3 short questions about my situation (accounts I have, budget, what I want the output to look like). 2. Propose a step-by-step plan using these tools or cheaper alternatives, with the time and cost of each step. Use official pricing pages; say "check pricing" if unsure. 3. Then do the setup with me one step at a time, waiting for my OK before anything that sends messages, posts publicly, spends money or deletes data. 4. Finish with a checklist to run it again next time.
You are helping me set up this AI use case end to end in Grok Bot. Goal: Build a hands-on Linux admin exam to test whether an agent can do your job What the original author did: Working on an agent to replace himself at work, a Linux admin built his own benchmark in the style of the RHCSA: one jumphost, three servers and 45 tasks, a third of them traps such as prompts hidden in log files. Reported result (as claimed by the author): A realistic test of agents on sysadmin work, with a full breakdown. Source: https://x.com/BlueTeamWarrior/status/2112027547764670729 Tools they used: Linux Category: Dev & ops Please: 1. Ask me 3 short questions about my situation (accounts I have, budget, what I want the output to look like). 2. Propose a step-by-step plan using these tools or cheaper alternatives, with the time and cost of each step. Use official pricing pages; say "check pricing" if unsure. 3. Then do the setup with me one step at a time, waiting for my OK before anything that sends messages, posts publicly, spends money or deletes data. 4. Finish with a checklist to run it again next time.
The Daily 5
Get five new use cases tomorrow morning
Free. Real sources. Unsubscribe any time.
Guides for this use case
Related use cases
Let Claude Code fix a Linux kernel bottleneck, then have humans review for races
Linus Torvalds let Claude Code fix a performance bottleneck in the Linux page cache. Kernel reviewers then found a race condition in the change, which shows why AI patches to core code still need careful human review.
Extend XFS into a clustered Linux filesystem with AI-written code
MeshXFS is an open-source clustered filesystem built on XFS. The original XFS code came from human developers; its author Stephen Shoecraft says AI wrote all the additions needed to make it clustered, and he did not review that code.
Write a Linux driver for an unsupported scanner by letting Claude watch the Windows driver
A GNOME developer wanted to scan his late mother's film negatives on a Plustek scanner that Linux no longer supported. He had Claude Code set up a Windows virtual machine with the scanner passed through, capture the USB traffic of the vendor driver, and compare it with Linux until scans worked. He then had Claude turn the scripts into a reusable hardware reverse-engineering harness.
Build a whole skeuomorphic Linux desktop environment with Claude writing the code
DEsk is a Wayland desktop environment for Linux with a skeuomorphic look. The author decided what it should be and how it should look and tested it on real machines, while much of the code and documentation was written by Claude under his direction.