Skip to content
Clairos Learn

Replayable training scenarios

Fix the system.
Build the instinct.

No command checklist. A starting condition, a concrete goal, and evidence that your repair worked.

Earn two practice stars for recovery and a third for diagnostic evidence. Hints never remove earned progress. Anonymous rewards last for this session. Signed-in practice rewards stay on this device and do not add competitive leaderboard XP.

Today’s incident

2026-09-06 · UTC

The site is down
01 · 5 min practice · Service recovery

The site is down

The training web service is not accepting requests. Inspect its state, bring it back, and verify the page responds.

Investigate
02 · 7 min practice · Safe configuration changes

The reload that failed

An invalid training configuration prevents nginx from starting. A known-good copy is available at /etc/nginx/nginx.conf.known-good. Inspect, restore, validate, then start the service.

Investigate
03 · 4 min practice · Least privilege

Your runbook is unreadable

Your private runbook is /home/learner/lab/private-note.txt. Recover owner-only read/write access, then read it. Do not grant group or public access.

Investigate
04 · 6 min practice · Evidence pipelines

Find the rejected logins

Inspect /var/log/training-auth.log. Save only denied events to lab/denied.txt and their newline count to lab/count.txt. Leave the original log intact.

Investigate
05 · 5 min practice · Artifact recovery

The missing homepage

nginx is active, but the root page is missing. A verified training copy is in /home/learner/backups/homepage.html. Restore /srv/www/index.html and verify it through HTTP.

Investigate
06 · 4 min practice · Operator communication

Leave a useful handoff

The runbook at /home/learner/lab/handoff-brief.txt states the incident name, observed state, and next action. Create lab/handoff.txt with those three exact key=value lines, keeping any credentials out.

Investigate
07 · 6 min practice · Verification after change

The changed release artifact

Compare lab/release.txt with backups/release-approved.txt. Restore only the changed working copy, then verify identical digests. Do not modify the approved copy.

Investigate
08 · 6 min practice · Evidence-led AI operations

A model is ready. The app is not.

Read /opt/ai-lab/incident.json. Identify the failing dependency and save its name followed by a newline in /home/learner/lab/dependency.txt. Model readiness alone does not explain application health.

Investigate

Earned through recovery

Repeat scenarios with fresh scratch state. The objective is safe understanding, not speed.