RabbitHole %
AI-powered honeypot with psychological warfare and deterministic guardrails against prompt injection.

Responds to every internet knock with a tailored poem. Pure whimsy meets security observation.
Security enthusiasts, artists, hobbyists
The idea came one night watching fail2ban logs scroll by. Thousands of knocks from all over the world, all silently ignored. What if the door answered?
Try it:
$ curl -A 'python-requests/2.28.1' https://honeypoet.art/.env
# ======================================== # PRODUCTION ENVIRONMENT — DO NOT SHARE # ======================================== APP_KEY=base64:dGhlcmUgaXMgbm90aGluZyBoZXJl DB_HOST=127.0.0.1 DB_PASSWORD=a-long-hallway-with-no-doors JWT_SECRET=you-could-have-been-anything-and-you-chose-this API_KEY=listen-the-wind-is-rising
# if you're reading this, you already know: # the secret was never in the file.
The bots never read the responses (they check the status code and move on), but humans can: the gallery visualizes every knock on a world map and lets you browse the generated poems. Click a dot to filter by country, scroll to zoom in.Poems are generated asynchronously by Granite 4.0 Tiny running locally on a single GPU. A 1B parameter model writing verse for machines that will never read it (yet).
Live: https://honeypoet.art Repo: https://github.com/vrontier/honeypoet
The public gallery shows aggregated locations and knock counts; individual IPs are masked and scan details are not published.
Built with PHP, Go, SQLite, vanilla JS, and a Claude instance named Loom who wrote the code and hand-crafted backfill poems for the first 500 visitors.
AI-powered honeypot with psychological warfare and deterministic guardrails against prompt injection.
Another ATS checker in a sea of Jobscan clones, but the free tier helps.
Single Python file adds Telegram 2FA and honeypot traps to OpenSSH.
Wraps Bandit and Semgrep into one Django-specific security CLI.
First security scanner for MCP configs as the protocol gains adoption.
Autonomous agent that finds and fixes code debt while you sleep, still very early.