Re:Show HN: Security Watch – alerts when a new vulnerability...
by VDudaG-A on Saturday November 21, 2026 @06:24PM
(#50594653)
Security Watch is an open-source tool (Apache-2.0) that tells you when a new security problem affects something you actually run, and where to look.
The problem: there is a lot of security news every day and almost all of it isn't about you. But sometimes it is, and then hours matter. A dependency scan in CI runs when someone pushes, and it only knows about the libraries in that repo. It doesn't know about the VPN box, the vendor product or the CI server your team runs. And often the first place a bug shows up is a vendor post or a news article, before there is a CVE with affected versions.
How it works: you give it a list of what you run. It can build that from your lockfiles (12 ecosystems), take SBOMs that your CI sends it, or you write entries by hand for things like a firewall or a vendor product. Each entry can have an owner and a note. Every 30 minutes it checks advisories (OSV, GitHub, cve.now, CISA's known-exploited list) and a few security news feeds, and when something matches you get a Slack message or email: which system, who owns it, which version, the fix if there is one. You decide what alerts at once and what waits for a daily digest.
News is matched by product name, so a finding from news alone is "possibly affected, please check", not a confirmed one. It reads the articles with plain rules, no LLM. One thing I learned doing that: articles say "no evidence of exploitation" about as often as they report exploitation, so wording matters a lot.
It runs on your own server. Your list stays there. Only package names and versions go out to the advisory lookups.
What it doesn't do: it doesn't scan your network or your running machines. It's only as good as the list you give it. News matching will miss some things and sometimes flag things you don't care about.
Install is `npm install -g @shieldaai/security-watch`, then `security-watch init --project ./your-app`. Node 24.10 or newer, or use the Dockerfile.
No account, no telemetry. I'd really like to hear from people who do this work every day: what it misses, and what would make the alerts worth reading.
The problem: there is a lot of security news every day and almost all of it isn't about you. But sometimes it is, and then hours matter. A dependency scan in CI runs when someone pushes, and it only knows about the libraries in that repo. It doesn't know about the VPN box, the vendor product or the CI server your team runs. And often the first place a bug shows up is a vendor post or a news article, before there is a CVE with affected versions.
How it works: you give it a list of what you run. It can build that from your lockfiles (12 ecosystems), take SBOMs that your CI sends it, or you write entries by hand for things like a firewall or a vendor product. Each entry can have an owner and a note. Every 30 minutes it checks advisories (OSV, GitHub, cve.now, CISA's known-exploited list) and a few security news feeds, and when something matches you get a Slack message or email: which system, who owns it, which version, the fix if there is one. You decide what alerts at once and what waits for a daily digest.
News is matched by product name, so a finding from news alone is "possibly affected, please check", not a confirmed one. It reads the articles with plain rules, no LLM. One thing I learned doing that: articles say "no evidence of exploitation" about as often as they report exploitation, so wording matters a lot.
It runs on your own server. Your list stays there. Only package names and versions go out to the advisory lookups.
What it doesn't do: it doesn't scan your network or your running machines. It's only as good as the list you give it. News matching will miss some things and sometimes flag things you don't care about.
Install is `npm install -g @shieldaai/security-watch`, then `security-watch init --project ./your-app`. Node 24.10 or newer, or use the Dockerfile.
No account, no telemetry. I'd really like to hear from people who do this work every day: what it misses, and what would make the alerts worth reading.