Semgrep

Code security for builders. Catch, flag, and fix real issues before they ship, powered by security that learns as you build.

[Semgrep @semgrep]

Following the Node2Shell and PHP2Shell incidents, we shipped a workflow to detect which repos may be affected: Detect-2Shell

How it works: It finds where the Node or PHP version is listed in Docker image tags, .nvmrc files, IaC, and CI setup steps. First, it finds version references in Python, it checks out whether that version is 🔴 vulnerable, 🟠 unpinned, 🟢 fixed, or ⚪ unknown. Then, it uses AI to look deeper, in places where the version is inside a variable, bundled image, or built from source. The results show where there is risk and guidance for how to fix it.

Detect-2Shell is now available for customers with Workflows enabled.

[Semgrep @semgrep]

If you're looking at an updated Node.js, and calling CVE-2026-111497 done? We have some bad news: We likely don't even know the true scale of "Node2Shell" yet.

Node patched CVE-2026-111497 on Oct 21. But the CVE wasn't in Node, it was in Brotli. So our researchers Diptendu Kar and Kurt Boberg went looking for every other package that bundles the same vulnerable Brotli decoder. Here's what we found

433 package entries across 12 ecosystems not great but the absolute worse part: Many of them don't even have fixes available yet. Now are they all reachable and exploitable? Probably not, we're finding out. But the important part: Node isn't the only affected runtime or package, not even close.

We're working through the list to manually verify each one but so far we're looking at a 71% no fix yet rate, of the fixes most are still in alpha/early releases.

The problem is most implementations aren't even called "brotli." The decoder is compiled into other runtimes like Chromium. 38 entries bundle Brotli older than 1.0.3. Some of those copies date to 2016, and of course are vulnerable to more than just this CVE. And there's no easy well to tell if you're affected.

For example, 19 are WebAssembly builds, where memory corruption exists, but stays inside the sandbox. Some require explicit configurations to enable Brotli compression and they aren't on by default. But it's not obvious unless you are looking for it and testing for it. We're going to keep validating (and producing rules), but it's going to be a rough weekend for incident responders.

[Semgrep @semgrep]

So while our agents worked on remediation, our Security Researchers broke down which Node.js apps can actually reach the bug: https://t.co/Nujl2vkFJ0

[Semgrep @semgrep]

With Node2Shell our security research team had details to share on our website. By the time we get to final review of a blog, a security check from our AI agent workflows flagged that we were running Node 23.11.0 too.

#AppSec #NodeJS

[Semgrep @semgrep]

But that's the speed security engineering has to keep up with now, and that's where our Gaurdian product was born. We needed to keep up with our software engineers using AI agents to build and be able to offload security investigations while our researchers worked with customers.

[Semgrep @semgrep]

Semgrep is a launch partner in the @spacexai Partner Network.

In Grok Build, a Hook runs Semgrep every time the agent changes code. Not when the agent remembers to, every time.

@spacerog on why that matters: https://t.co/f26Ary2OFD

[Semgrep @semgrep]

How does the vulnerability work? Read the technical analysis by our Security Researcher, Kurt Boberg: https://t.co/IslFuBev2n

[Semgrep @semgrep]

PHP2Shell is a critical vulnerability affecting 10 years of PHP releases. Semgrep's Guardian Plugin for Claude Code and Codex now flags if your PHP installation is vulnerable as soon as your agent generates PHP code.

[Semgrep @semgrep]

“You can’t YOLO at enterprise scale.” @daghanaltas at #GartnerSYM this morning. Yesterday we said to come argue with him. Several people took us up on it.

[Isaac Evans @0xine]

If Claude Code is the new us-east-1 (I think it is), your coding agents should know about php2shell before your cloud scanner.
@Semgrep’s Guardian plugin for Claude Code, Codex & more now deterministically flags vulnerable PHP versions. Nice work team!

[Semgrep @semgrep]

PR volume: up 3.1x. Reviewers: up 1.5x. PRs getting any written feedback: 39% → 21%.

Tomorrow at Gartner IT Symposium, @daghanaltas on what replaces reviewing every change by hand. Stage 3 (Pacific), 10am ET.

Can’t make it: https://t.co/zOc3y2q3x6

[Space Rogue @spacerog]

Your company has security tools, policies, scanners, and enough dashboards to be a TV studio. Congratulations.

But will any of it actually work when something goes wrong?

Part 3 of my series on why we need to rename it to Cybersecurity Readiness Month.

https://t.co/bSdsK0VLwF

[Semgrep @semgrep]

Day 2 at Gartner. Edwin, our PMM: “There’s a solution for pretty much everything. What people are missing is a standard around how to secure AI.”

[Semgrep @semgrep]

AI coding agents are generating more code, faster. Security needs to keep up. 🔒

Join us for a hands-on session with Semgrep Guardian to see how security fits directly into AI coding workflows with Claude and Codex.

See how Guardian detects vulnerabilities as code is being written and helps teams address security risks earlier, without slowing developers down.

🧱 Build beyond the code! Join live for a chance to win one of three $50 LEGO gift cards.

📅 October 21 | 8 AM PT

Get hands-on with Guardian → https://t.co/juPjc2M2yy

[Semgrep @semgrep]

Your developers are shipping faster than ever. Your security team is trying to keep up.

On Oct 22, Isaac Evans (CEO, Semgrep) joins Viraj Gandhi from ServiceNow and Varun Agaskar from Broadridge Financial for the session “Security at the Speed of Development” at the PBC Agentic AI Security Conference.

It’s a conversation, not a slide deck. Here’s what they’ll get into:

How the secure SDLC changes when AI writes more of the code
What it takes to build security into developer and agent workflows, not just bolt it on at review time
Which policies can become automated checks, and which still need a human call
Why application and infrastructure risk belong in the same conversation

Join us for real talk and practical steps you can take tomorrow.

October 22, 10:00 - 10:30 AM PT

ttps://armorcodeinc.registration.goldcast.io/events/0a66b331-f6fb-4bd2-897f-468a7c6ed390?__hstc=14956073.22adf95f00a7cb40122d709247769689.1791980575910.1791980575910.1791980575910.1&__hssc=14956073.1.1791980575910&__hsfp=a0a1851a45f4a97708ee74c66b52c7bc#Registration

[Semgrep @semgrep]

They did not break in: they published.

That line from the Beagle Brigade team sums up how package registry malware works. A stolen maintainer token becomes a new version of a package you already depend on, and it runs on every laptop and CI runner that installs it, before anyone opens a file.

Their project took second place in our Semgrep Guardian challenge at the Cyberdefense Hackathon during SF Tech Week.

Each new release on a registry like npm or pypi is pulled into a sandbox. It’s analyzed with Semgrep dataflow rules without ever being executed. Anything suspicious goes to an AI agent that returns a verdict.

What we liked most is that the analysis never runs what it downloads. When the threat is code that executes on install, analysis that doesn’t execute anything is the safe place to start. It’s the same threat Semgrep Guardian’s malware firewall is built for with malicious dependencies.

Congrats to the Beagle Brigade team!

#SFTechWeek

[Semgrep @semgrep]

Repos hosted in Cursor Origin can now connect to Semgrep. Automated scans, PR checks, and findings in the workflow you already use.
https://t.co/OSusjI9AEt

[Semgrep @semgrep]

Only raise a security issue you can prove.

That was the idea behind a hackathon project which took first place in our Semgrep Guardian challenge at the Cyberdefense Hackathon during SF Tech Week.

A static deterministic check runs first, so most saves never reach a commit at all. The findings go into the model's context as evidence. Anything the model suspects is then run in a sandbox with canaries. Only proven issues reach the developer, and each comes with a fix that’s re-checked the same way.

Static analysis is cheap and precise about what the code says. A model is good at asking whether it matters in a project. Of the 63 projects submitted, this was one of the clearest examples of each security layer doing the task it’s best at.

Congrats to the Reva team!

#SFTechWeek

[Cursor @cursor_ai]

Origin, Cursor’s code hosting platform, now supports over 20 integrations.

Connect your repos to tools for CI, deployments, code review, security, monitoring, and project management.