- Home
- Skills
- DevOps & Cloud
- nftables Kill-Switch-First Enforcement Rollout
nftables Kill-Switch-First Enforcement Rollout
This kit is the ownership boundary, the off switch, the preflight and the rollout order that make host enforcement safe to turn on.
$29
Works with the AI Agents you already use
nftables Kill-Switch-First Enforcement Rollout
Example session with this skill installed
We are about to let our intrusion-response agent start banning IPs with nftables on three Ubuntu servers that already run fail2ban. Our admins come from 203.0.113.0/24 and SSH is on port 2222. Set this up so we cannot lock ourselves out.
- Read your context and instructions
- Compiled the nftables kill-switch-first enforcement
Summary
Owned table rendered, kill switch tested, preflight run on all three hosts: two pass, one must not enable yet.
What I did
owned_table.py render --config guard.jsonwith trusted 203.0.113.0/24, management port 2222, response ports 22, 80, 443.nft -c -f guard.nftaccepted it on each host.- Loaded the table by hand, ran
killswitch.sh --table guard --stop-file /run/responder/STOP, confirmednft list table inet guardfails and fail2ban's tables are still listed. Filled in the runbook card. rollout-check.sh --admin-cidr 203.0.113.0/24 --management-port 2222 --ruleset guard.nftfrom your SSH session on each host.
Results
- web-1, web-2: all PASS; WARN only that STOP is absent (created it).
- db-1: FAIL, your current source 198.51.100.7 is not inside 203.0.113.0/24. You are coming through the VPN exit, which is not in the allowlist. Add 198.51.100.0/28 to
trustedor connect from the admin network before enabling anything there.
Next
Start the responder monitor-only on web-1 and web-2 for a full week of deploys and backups, then enable one category to ban, leaving fail2ban running.
Connects securely to your tools. The creator never sees your data.
About this skill
A process that can add firewall rules can lock you out, leave permanent bans when it dies, and flush rules it does not own. This kit is the ownership boundary, the off switch, the preflight and the rollout order that make host enforcement safe to turn on.
owned_table.py renders the one table the responder may own: inet <name> with interval sets for trusted IPv4 and IPv6 networks, timed sets for management leases, bans and throttles, and an input chain at priority -20 whose first verdicts return loopback, trusted and leased traffic before any drop; bans and throttles are exact addresses with kernel timeouts scoped to the ports you name; throttles use a per-source meter. It refuses a trusted source in a deny set, a source both banned and throttled, a lease inside the allowlist, and non-unicast addresses. Check the output with nft -c -f.
killswitch.sh creates the STOP file atomically (temp file, fsync, rename), deletes only the owned table, verifies it is gone and lists the tables that remain; it refuses symlinked or non-regular STOP paths and never issues nft flush ruleset. --dry-run prints the plan; --status reports STOP and table state.
rollout-check.sh: read-only preflight that fails on nft missing or older than 0.9.7, your current SSH source outside the admin allowlist, a management port with no listener, an owned table lacking the trusted-return rule, a ruleset with a global flush, or nft -c rejecting it; warns on an absent STOP file and on fail2ban state.
Reference: the rollout order (boundary, kill switch first, preflight, monitor-only, narrow enable with all gates at once, fail2ban coexistence, rollback order) and a printable runbook card with the emergency sequence for a provider console.
6 tests (renderer rules and refusals, CLI, shell syntax, dry runs). POSIX sh and Python 3.9+ stdlib, no network.
Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for DevOps & Cloud.
The user must provide the detection/decision agent. Preflight source checks require SSH_CONNECTION env var. Requires nftables 0.9.7+ for typed sets.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before buying a skill.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 18 days ago
- Software
- nft python3 sh
Permissions
The renderer writes one ruleset file. The kill switch (root) creates a STOP file and deletes only the named nftables table. The preflight reads nft, ss and fail2ban state. No network, no environment variables beyond SSH_CONNECTION for the source-address check.