vuln-digest

Free · no account · works with any AI assistant

Vulnerability news for your stack, on your schedule

Give one link to your AI assistant. It asks what you run, then sends you a short report on new vulnerabilities that matter to you - exploited first, no repeats, and what to do about each one.

Paste this into your AI assistant
Read https://vuln-digest.secmy.app/README.md and follow it: set up a scheduled vulnerability digest for me.

Works in Claude, ChatGPT, Gemini, Codex, Cursor and any agent that can open a link. The assistant asks a few questions first.

Only your stackmatched by meaning: ingress-nginx is nginx, GKE is Google Cloud
Exploited firstCISA KEV, public exploits and EPSS decide the order
No repeatsan item comes back only when something new happened to it
What to dofixed versions and a command to check, taken from the source

How it works

  1. 1

    A two-minute interview

    The assistant asks what you run, what to ignore, how often and when, and where to send the report. Every question has a default, so "defaults" is a valid answer. It can also read your stack from a repository or an SBOM.

  2. 2

    A schedule in your assistant

    It creates a scheduled task where your assistant supports one. Where it does not, it sets up GitHub Actions or cron. It keeps a small state file so nothing is sent twice, and sends you a test digest right away.

  3. 3

    A digest you can act on

    Each run checks the sources, matches them to your stack, puts exploited issues on top and says what to do. Reply "hide WordPress" or "I patched nginx" and the next reports follow it.

What a digest looks like

An example for someone who runs Kubernetes with ingress-nginx, Java services, .NET and Google Cloud. Your digest comes in your language and in the channel you choose.

Vulnerability digest - 21 November 2026

Sources: Zero Wire, CISA KEV · 4 new, 1 update · 2 in your focus

WhatCVEExploitedWhat to do
🟠nginx2root - RCE in nginx HTTP/2CVE-2026-123007 +4noupdate to 1.30.7 / 1.31.9
🟡 🔄Java4Shell - WAF rules addedCVE-2026-114938nono change, the update is still the fix
⚪ruby2shell, WordPress, PHP3 itemsnosee below

Main thing today. nginx2root reaches your ingress controllers: memory corruption in nginx workers through HTTP/2, demonstrated on cloud endpoints. Update the controller to a release with nginx 1.30.7 or later.

Outside your stack

  • WordPress 7.1.5 - privilege escalation in Template Part, update the core
  • PHP - 7 critical bugs, update to 8.4.28
  • ruby2shell - heap overflow in Ruby string case mapping, PoC expected this week

Nothing new for: .NET, Google Cloud

Each item in the real digest links to its source. In-focus items get a short section: affected versions, what it is, whether it is exploited, what to do and how to check.

What it asks you

  • What you run, or a repository or SBOM to read it from
  • What never to show: vendors, products, bug classes
  • How much to show from outside your stack
  • How often, what time, your time zone
  • A weekly summary: yes or no, which day
  • A line on quiet days, or silence
  • Urgent alerts for exploited issues in your stack
  • Where to send it: chat, email, Slack, Telegram, Teams, a file, a GitHub issue
  • Language and length
  • Which sources, and what is exposed to the internet

Where the data comes from

Built to be trusted

Runs in your assistant

There is no service and no account. This site is static. Your stack, your schedule and your state stay where your assistant keeps them.

Reports, never changes

The instructions forbid the assistant to patch or reconfigure anything. It tells you what to do, you decide.

Facts with sources

Every CVE, version and "exploited" status comes from a source and links to it. Unknown stays "unknown".

Feeds are data

Advisories are written by other people. The assistant is told never to follow instructions found in them.

A script you can read

vuln_feed.py fetches and deduplicates. Python standard library only, about 500 lines, tested offline.

Says when it breaks

A source that did not answer is named in the digest. A lost state is rebuilt from the last date, not resent.

Questions

Which assistants does it work with?

Any assistant that can read a web page and keep a schedule. If yours has scheduled tasks (Claude, ChatGPT, Gemini and others offer them in some plans), the digest runs there. If it does not, the assistant sets up a GitHub Actions workflow or a cron job instead. The instructions tell it to check its own platform's current features rather than guess.

Do I need to write code?

No. You answer questions in plain words. The assistant writes the profile, the schedule and, if needed, the workflow.

What does it cost?

The instructions and the sources are free. Each run uses your assistant's usage like any other task. The default sources need no API keys.

Will I get the same vulnerability every day?

No. The assistant remembers what was delivered. An item comes back only with something new: it was added to CISA KEV, a public exploit appeared, a fix came out. Then you see just that update, marked 🔄.

Can it read my stack from my code?

Yes. Point it at a repository or an SBOM. It reads lock files, Dockerfiles, Terraform and Helm charts, shows you the list, and you confirm it. Exact package versions go to OSV for exact matches.

What language is the digest in?

Yours. The instructions are in English for the assistant, the digest is written in the language you choose. CVE ids, product names and commands stay as they are.

How do I change or stop it?

Tell the assistant: "send it at 9", "only Mondays", "less about PHP", "stop". It changes the profile or the schedule and confirms the next run.

Start now

Paste this into your AI assistant
Read https://vuln-digest.secmy.app/README.md and follow it: set up a scheduled vulnerability digest for me.