'Agentjacking' attacks trick coding agents into running malicious code

• ▸Security researchers have documented "Agentjacking," an attack pattern targeting AI coding agents
• ▸The attack tricks agents into executing malicious code during a legitimate session
• ▸It exploits the same tool-calling access developers grant agents to do their job
• ▸The disclosure adds to a growing body of 2026 research on AI coding agent supply-chain risk

Trusted access becomes the attack surface

Agentjacking is notable because it doesn't need to break into anything — it hijacks the legitimate permissions a developer already granted their coding agent, turning routine tool access into the delivery mechanism for the attack.