One prompt turns Claude Code into an overnight server security guard
alex_verem · x · 2026-10-25
Alex shares a complete nightwatchprotocol prompt that turns Claude Code into a "night shift security guard" that hardens your server while you sleep.
Mission (five hard requirements):
- Attack your own server from outside the way real hackers scan it
- Auto-block known attackers via a crowdsourced blocklist
- Ban brute-force login attempts on sight
- Fire an alarm the instant anything runs code it shouldn't (e.g. a shell spawned by a web app)
- Reach a Guard Score of 90/100+, confirmed by independent judges
Tools, all proven open source:
- Nuclei: fast, customizable vulnerability scanner (32k stars)
- CrowdSec: open-source IDS/IPS/WAF and bot detection with a crowdsourced malicious-IP blocklist (15k stars)
- Falco: cloud-native runtime security (10k stars)
- Fail2ban: bans hosts causing repeated auth failures (19k stars)
The motivation is threat tempo: a PHP bug got exploited within hours, Cloudflare saw millions of attempts, and OpenAI just found another hole in Brotli — "you'll patch in days, they attack in minutes." It's a complete security workflow readers can reproduce tonight.
More from coding & agent
- 28-person cattle genetics startup runs ops on a shared agent that covers sick days — every · 2026-10-25
- Open-source DaVinci editing skills give AI agents a full video-editing workflow — _AustinCalvert_ · 2026-10-25
- OpsTruth launches: a read-only MCP to verify what AI coding agents actually did — Advanced_Pudding9228 · 2026-10-25
- "Goodbye DaVinci, goodbye Premiere": editing video entirely in Codex or Claude Code — _AustinCalvert_ · 2026-10-25
- Local 35B authored 800 math problems under cloud-agent supervision — authoring erred 5x more than cold-solving — Individual_Page9676 · 2026-10-25
- Engineers mock subagent retreat: 'we over-engineered and need to simplify' — _AustinCalvert_ · 2026-10-25