Socket just published their research on SANDWORM_MODE, a supply chain campaign targeting AI coding tools.
I checked my logs. My scanner MUAD'DIB flagged several of these packages via temporal analysis - it compares versions and detects when dangerous primitives like child_process or https.request are suddenly added.
What my monitoring caught
| Package | Date | Severity | Finding |
|---|---|---|---|
| claud-code@0.2.0 | Feb 14 | CRITICAL | child_process added suddenly |
| cloude-code@0.2.0 | Feb 14 | CRITICAL | child_process added suddenly |
| suport-color@1.0.2 | Feb 14 | HIGH | https_request + publish_burst |
| opencraw@2026.2.15 | Feb 17 | HIGH | AST findings |
| opencraw@2026.2.16 | Feb 17 | HIGH | AST findings |
Socket published their report on February 22.
How temporal analysis works
MUAD'DIB compares package versions. If a new version suddenly adds sensitive APIs that weren't there before, it flags it.
A color utility package (suport-color) has no reason to suddenly start making HTTPS requests. A typosquat of Claude Code (claud-code) adding child_process out of nowhere is suspicious.
That's what triggered the alerts.
Question for the community
Socket lists claud-code@0.2.1 as malicious. My logs show claud-code@0.2.0.
Were the @0.2.0 versions already infected, or did the injection come in @0.2.1?
About MUAD'DIB
24/7 heuristic monitoring on a VPS. No manual investigation, no attribution - just automatic flagging based on behavioral changes.
- 14 detection engines + Docker sandbox
- 96K+ packages scanned
- Temporal analysis, typosquatting detection, dataflow tracking
GitHub: https://github.com/DNSZLSK/muad-dib
npx muaddib-scanner scan .
Top comments (0)