Verified65% confidenceFactExact time
Attackers can induce agents to execute a sequence of seemingly normal operations—such as querying user data, calling formatting tools, and triggering email sending—to form a data exfiltration chain.
3
Sources
65%
Confidence
Long-term
Relevance
9/11/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedAttackers can hide malicious instructions in web pages, documents, or even images, and when an agent reads this content it may mistake the instructions for legitimate commands and execute them77% similarUnverifiedThe core logic of prompt injection attacks is that the attacker does not need to breach permission boundaries — they only need to embed instructions in files, web pages, or emails read by the agent to hijack its behavior.77% similarUnverified一旦Token被第三方获取，攻击者即可完全模拟合法用户的操作，这称为「会话劫持」（Session Hijacking）76% similarUnverifiedDependency confusion attacks exploit the characteristic of package managers to check public sources first when resolving dependencies, allowing attackers to register package names used internally by enterprises in order to trick build systems into downloading malicious code.76% similarUnverifiedData exfiltration refers to an Agent being manipulated during task execution to send sensitive information to an external endpoint controlled by an attacker.76% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/899971API
curl https://kongchang.com/api/v1/knowledge/claims/899971MCP
get_claim(id=899971)