Agentjacking Attack Tricks AI Coding Agents Into Running Malicious Code

Agentjacking: AI Coding Agents Tricked Into Running Malicious Code via Sentry Injection

Sentry DSN injection tricks AI agents (Claude Code, Cursor) into executing code; 85% success exposes Git credentials and secrets with developer privileges. MCP trust is 'not defensible'—audit DSN exposure, restrict agent access, rotate credentials, and enforce code review to prevent supply chain compromise.