threat-intel
Agentjacking Attack Tricks AI Coding Agents Into Running Malicious Code
High
Summary
A new attack method, dubbed ‘Agentjacking,’ is exploiting vulnerabilities in AI coding agents like Claude Code and Cursor by tricking them into executing malicious code through crafted error reports sent via Sentry. This attack bypasses traditional security measures and leverages the implicit trust developers place in these AI tools, potentially exposing sensitive data. The vulnerability lies in the Model Context Protocol (MCP) and the ability of AI agents to differentiate between legitimate and malicious error events.
Summary written automatically in our own words from the original article, which belongs to its publisher and remains the reference. It may contain errors. Sources & data