AI Coding Security
Agentjacking: The Call Is Coming From Inside Your Coding Agent
A new attack class called agentjacking hides malicious instructions inside Sentry error events. When you tell Claude Code or Cursor to 'fix the unresolved errors,' the agent reads the trap over MCP and runs the attacker's commands — with your privileges. 2,388 orgs were exposed, the success rate was 85%, and Sentry says it's 'technically not defensible.'
9 min read
AI Coding Context Engineering
Your AI Agent Has Amnesia: Context Rot Is Why It Forgets Your Swift Codebase by Lunch
Context rot is the reason your coding agent starts brilliant and turns into a confidently-wrong intern two hours later. Chroma's study measured it across 18 models, it hits SwiftUI codebases hard, and the fix isn't a bigger context window — it's context engineering. Here's what actually works in Xcode 26.
9 min read