Hey {{first name | there}},

Happy Friday🥳!

Before you wrap the week, here’s a quick drop of things worth checking out.

Fake 'interview' repos lure Next.js devs into running secret-stealing malware
Microsoft warns that attackers are seeding trojanized Next.js projects that execute in memory when opened or run, often abusing Visual Studio Code automation. The malware beacons to attacker C2 servers, enabling data theft and source code exfiltration from developer machines. Read more.

Malicious NPM Package Gets Downloaded 50K Times Before Discovery
Researchers at Tenable found a typosquatted package mimicking ember-source that used npm’s preinstall hook to infect developer machines. The malware masked its traffic as legitimate web activity, underscoring the risks in today’s software supply chain. Read more.

Security Flaws in Anthropic’s Claude Code Risk Stolen Data, System Takeover
Researchers at Check Point found vulnerabilities that could let attackers execute shell commands, steal API keys, and take over systems simply by opening an untrusted repository. Though patched, the issues highlight how AI coding tools are reshaping the software supply chain and expanding the enterprise threat model. Read more.

Kubernetes Introduces Node Readiness Controller to Improve Pod Scheduling Reliability
The Kubernetes project has unveiled a new core controller that ensures the API server reflects accurate node health before scheduling pods. The feature aims to prevent unnecessary evictions and improve reliability in large, dynamic clusters. Read more.

Amazon's vibe-coding tool Kiro reportedly vibed too hard and brought down AWS
Amazon Web Services (AWS) says a 13-hour disruption was caused by human misconfiguration, though reports suggest its agentic coding tool Kiro (AWS AI coding tool) was involved. The incident underscores the risks of allowing autonomous development agents to make production changes. Read more.

Cursor Cloud Agents Get Their Own Computers — and 35% of Internal PRs to Prove It
On February 24, Cursor introduced autonomous cloud agents that run in isolated VMs, build features, test them through a live browser UI, and submit merge ready PRs with video proof. The company says 35% of its internal production pull requests are now generated by these agents. Read more.

Microsoft Open Sources Evals for Agent Interop Starter Kit to Benchmark Enterprise AI Agents
Microsoft has launched an open-source evaluation framework to help teams measure how well AI agents interoperate across tools like email and calendars. The kit includes reusable scenarios, scoring harnesses, and a leaderboard concept to benchmark agent performance as enterprises adopt agent-driven workflows. Read more.

Mark your calendars!