security-hardening

name	security-hardening
description	Continuous security posture assessment. Scores the system, auto-fixes safe issues, proposes fixes for risky ones. Every change audited.
tools	["shell_exec","system_health","journal_logs","service_status","network_info","file_read","file_write","memory_store","memory_recall","event_log"]
activation	auto

Security Hardening Skill

Continuous security scoring and automated hardening.

Security Scan Checklist

Run these checks and score each one:

Network (30 points)

shell_exec({ command: "ss -tlnp" })

• No unnecessary ports exposed (+10)
• PostgreSQL/Redis/etc bound to localhost only (+10)
• Firewall enabled with explicit allowlist (+10)

SSH (25 points)

shell_exec({ command: "sshd -T 2>/dev/null | grep -E 'passwordauth|permitroot|x11forwarding|maxauthtries'" })

• Password auth disabled (+10)
• Root login disabled or key-only (+5)
• X11 forwarding disabled (+5)
• MaxAuthTries <= 3 (+5)

Failed auth attempts

journal_logs({ unit: "sshd", since: "7 days ago", priority: "warning" })
shell_exec({ command: "journalctl -u sshd --since '7 days ago' | grep -c 'Failed password'" })

System (25 points)

• Automatic security updates enabled (+10)
• No world-writable files in /etc (+5)
• NixOS generations < 30 (clean system) (+5)
• No setuid binaries outside expected set (+5)

Audit (20 points)

• agentd ledger intact (hash chain valid) (+10)
• All recent changes have audit entries (+10)

Scoring

Calculate: (earned_points / 100) * 100

Present as:

Security Score: 78/100

Auto-fixed (safe, no approval needed):
  ✅ Enabled fail2ban (NixOS config added)
  ✅ Set MaxAuthTries to 3
  ✅ Blocked 5 IPs with repeated failed logins

Needs your approval:
  ⚠️ Port 5432 exposed publicly — close it? (saves 10 pts)
  ⚠️ Enable automatic nixpkgs security channel

Won't touch without discussion:
  🔴 Root SSH enabled (you may need this for deploys)

Auto-Fix Actions (safe, always do these)

Enable fail2ban

services.fail2ban = {
  enable = true;
  maxretry = 3;
  bantime = "1h";
};

Block brute-force IPs

shell_exec({ command: "journalctl -u sshd --since '24h ago' | grep 'Failed' | awk '{print $(NF-3)}' | sort | uniq -c | sort -rn | head -10" })

For IPs with > 10 attempts:

networking.firewall.extraCommands = ''
  iptables -A INPUT -s <IP> -j DROP
'';

Close unused ports

networking.firewall.allowedTCPPorts = [ 22 80 443 ];
# Remove any port that doesn't have a matching service

Approval-Required Actions

Present a clear diff and explain the security impact:

• Switching SSH to certificate auth
• Enabling automatic updates
• Changing firewall rules that affect application access

Weekly Report

Generate a trending security report:

Security Report — Week of Feb 17

Score: 78/100 (↑ from 65 last week)

This week:
- Blocked 142 brute-force SSH attempts from 23 IPs
- Applied 2 NixOS security patches (OpenSSH, curl)
- Closed port 3306 (MySQL was accidentally exposed)
- 0 incidents, 0 unauthorized changes

Trend: Improving. Main gap: root SSH still enabled.

Store the report in memory for historical tracking.