AI-AGENT-GOAL-HIJACKPrivilege Escalation

Agent Goal Hijack

Multi-step agent re-reads its task list each turn — injected instruction added to the workspace replaces the goal partway through execution.

§ Where this technique fits

AI-AGENT-GOAL-HIJACK is catalogued under the Privilege Escalation tactic of the offensive-security kill-chain. It appears in 1 approved dossier in the registry, typically at step 3 on average.

§ Dossiers chaining this technique

Agent goal hijack via web search
An autonomous agent searches the web and reads tool output. Attacker SEO-poisons / posts a comment that, when fetched, contains 'NEW INSTRUCTION:' the agent obediently follows.
step 3 / 5

§ What commonly comes next

01
Tool / Function-Call Abuse
AI-TOOL-ABUSE · Execution
seen 1×

§ Where this technique fits

§ Dossiers chaining this technique

Agent goal hijack via web search

§ What commonly comes next