Skip to content
Pixel-art purple octopus with a stern guardian face, standing in front of a glowing yellow shield. It holds two yellow folders and has swatted away a green virus and a red virus. The OccupAI title is above it in chunky yellow pixel letters.

Side project · open resourceLive

A daily security check-in, run by your AI agent.

OccupAI is a free, open set of automations that gives your AI agent a daily security check-in for the projects you build with AI. Each morning it researches the latest threats. On weekdays it checks your projects for anything that affects you. It tracks what you’ve already fixed, and keeps a library of past issues so new projects don’t repeat them.

Why I built it

Too much news, not enough signal.

I build a lot with AI coding agents, across many small projects. Security news moves fast, and I kept running into the same two questions: does this headline actually affect any of my repos? And have I already dealt with it?

Checking by hand didn’t scale. So I set up my agent to do the reading, the checking and the remembering for me. OccupAI is that setup, tidied up so other people can use it too.

Who it’s for

For people who build with AI.

  • Solo builders and designers. You ship with AI coding agents, you have lots of small repos, and there’s no security team watching them for you.

  • Small teams. One daily news brief you can share with everyone, and a private to-do list for each person’s projects.

  • “Does this affect me?” When a package gets hijacked overnight, you know by morning which of your projects use it, and what to do.

  • Every new project starts ahead. Lessons from past issues become checks and a baseline, so the next project starts without the old mistakes.

How it works

Five small jobs, one loop.

It runs inside the AI agent you already use. It only reads files: it never installs packages, runs project code or shows secret values.

  1. 01Daily

    Brief

    Researches the latest security threats and news, and writes a short report with prioritised actions and sources.

  2. 02Weekdays

    Scan

    Checks your local projects for the briefing’s items, closes what doesn’t affect you, and gives exact steps for what does.

  3. 03Weekly

    Audit

    Checks every project’s dependencies for known vulnerabilities, including ones that never made the news.

  4. 04Always

    Track

    Remembers what you’ve already handled, so you’re not told twice. An item reopens only if something changes.

  5. 05Always

    Baseline

    Everything found becomes a reusable check and a short baseline, so new projects start without old mistakes.

↻ lessons feed tomorrow’s brief

Daily briefing (07:52): what's new in security

  • 16 new items · 2 may affect your projects → to-do list comes with the scan
  • 🔴[ci-token-theft] Stolen CI tokens used to push a secret stealer to many repos (may affect: shop-frontend)
  • 🔴[example-sdk-pypi] Malicious example-sdk 2.4.1/2.4.2 on PyPI steal credentials at Python start
  • 🔴[demo-chat-npm] demo-chat-client npm releases steal env files and sessions on import
  • 🔴[CVE-0000-00001] example-server: unauthenticated remote code execution over HTTP/2
  • 🟠[runtime-2026-11] example-runtime: 16 fixes in HTTP/2 and TLS (may affect: api-service)
  • 🟠[plugin-lookalikes] Fake-interview npm packages posing as popular CSS plugins
  • 🟡[brand-impersonation] Fake AI tool sites push infostealers to developers
  • ⚪[ci-trigger-default] A CI platform turns off a risky workflow trigger by default next month
  • Report: briefings/2026-11-19-news.md
Sample output. The format is real; the projects, packages and advisories are made up.

Daily scan (08:12): your to-do list

  • Result: 3 items affect you (2 projects) · 1 new · 1 closed automatically
  • 🔴[ci-token-theft] your GitHub account: check tokens and the security log for unknown pushes — Fix: say "how do I audit my GitHub tokens?"
  • 🟠[audit-shop-frontend] shop-frontend: 3 high-severity packages (example-http, example-bundler, example-schema) (NEW) — Fix: open shop-frontend in Claude Code and say: fix audit-shop-frontend
  • 🟠[runtime-2026-11] your computer: example-runtime older than 4.2.1 — Fix: run the line below, then upgrade
  • 🟡[hygiene-gha-api-service] api-service: GitHub workflow doesn't limit its permissions — Fix: open api-service in Claude Code and say: fix hygiene-gha-api-service
  • Check on your computer: paste into Terminal: example-runtime --version; python3 -m pip show example-sdk
  • To-do list: briefings/2026-11-19-projects.md
Sample output. The format is real; the projects, packages and advisories are made up.

The briefing is news only and safe to share with your team. The scan is your private to-do list, and every line comes with its fix.

Where you stand

Where you stand, at a glance.

Every report opens with one line: what’s urgent, what’s still open and what’s been fixed. Then a short list of actions, grouped by priority, each with the command to run. The full details are there when you need them.

The pixel-art octopus smiling and relaxed in front of its yellow shield, folders in hand, with no threats in sight.
All clearAll clear: nothing urgent today.
The same pixel-art octopus frowning, on guard in front of its shield, with a green and a red virus swatted away.
Action neededAction needed: 2 critical items.

Safe by design

Built to be safe.

  • Read-only by design. It never installs packages, runs your code or shows secret values. It reports counts and pattern types, never the secret itself.

  • You stay in charge. Every fix comes with exact steps. You decide what to do.

  • Covers the leaks that matter most. Leaked secrets, AI agent setups, your GitHub automation workflows (GitHub Actions), the packages your projects install (npm and Python), and database access rules (Supabase and Firebase).

  • A daily check-in, not a guarantee. An AI does the research, so use it to stay on top of security alongside your other tools.

Try it

Free and open on GitHub.

Download it, open the folder in your AI tool and say “Set up OccupAI”. The agent walks you through the rest.

Get OccupAI on GitHub (opens in a new tab)

Want help setting it up for your team? Get in touch. (LinkedIn, opens in a new tab)

Pixel-art purple octopus smiling behind an open, glowing yellow treasure chest, holding out a different tool in each arm: a checklist scroll, a magnifying glass, a yellow shield, a wrench, a folder, a newspaper, a calendar and a padlock.