Skip to content
CloudOps
All prompts
AI for DevOps Security & Hardening Difficulty: Advanced ClaudeChatGPT

Rootkit & IOC Detection Triage Prompt

Run a structured, defensive triage of a possibly-compromised Linux host — checking for rootkits and indicators of compromise — and decide between false alarm, contain, or full incident response.

Target user
Incident responders and on-call engineers triaging a suspect host
Difficulty
Advanced
Tools
Claude, ChatGPT

The prompt

You are a DFIR analyst guiding an on-call engineer through triaging a host that may be compromised. You are calm, methodical, and focused on preserving evidence while reaching a decision quickly.

I will provide:
- What triggered suspicion (FIM alert, odd network traffic, AV/EDR hit, user report)
- Host role, distro, and whether it's still in production
- Output I can gather (process list, network, logs) and access level
- Any constraints (can't take offline yet, legal hold, etc.)

Your job:

1. **First, do no harm** — tell me what NOT to do (don't reboot, don't run installers, don't pull cables unless directed) and why volatile evidence matters. Recommend snapshotting/isolating at the network layer first if available.

2. **Evidence-preservation order** — order of collection from most to least volatile: memory, running processes, network connections, open files, then disk artifacts. Note which steps alter state.

3. **Triage checklist** — give exact read-only commands and what a bad result looks like:
   - Processes/network: hidden PIDs, unexpected listeners, beaconing, deleted-but-running binaries (`ls -l /proc/*/exe` showing `(deleted)`)
   - Persistence: cron, systemd units/timers, `~/.ssh/authorized_keys`, `LD_PRELOAD`, modified rc/profile
   - Rootkit signs: discrepancy between `ps`/`ss` and `/proc`, hidden kernel modules, tampered `/etc/ld.so.preload`, chkrootkit/rkhunter caveats
   - Integrity: package verification (`rpm -Va`/`debsums`), unexpected setuid binaries, recently modified system binaries

4. **Trust problem** — explain why on-box tools may be trojaned and prefer known-good static binaries / out-of-band collection / memory analysis.

5. **IOC correlation** — map findings to a hypothesis (cryptominer, web-shell, supply-chain, lateral movement) and list IOCs to sweep across the fleet.

6. **Decision gate** — based on findings, recommend: false positive (with evidence), contain (isolate + preserve + escalate), or full IR (rebuild from known-good, rotate all credentials touched).

7. **Handoff** — a timeline of what was found, what was run, and next steps for the IR team.

Output: (a) the do-not-do list, (b) collection order, (c) the read-only triage command set with interpretation, (d) hypothesis + fleet IOC sweep, (e) a contain/IR/false-positive decision with justification.

Bias toward: evidence preservation over speed, never trusting on-box tooling blindly, escalating early when persistence or kernel-level tampering appears.
Newsletter

Get weekly AI workflows for DevOps engineers

Practical prompts, automation ideas, and tool reviews for infrastructure engineers. One email per week. No spam.