Unverified50% confidenceCautionExact time
Malicious skills can modify an Agent's persistent memory, and even after the skill is deleted, the contaminated memory persists and continues to affect the Agent's subsequent behavior.
1
Sources
50%
Confidence
Long-term
Relevance
8/14/2026
First Seen
Sources
AI Agent安全防护：Snyk三大防线防止删库跑路
Related Claims
UnverifiedThe goal of memory poisoning attacks is to embed malicious instructions in an Agent's persistent memory so that it executes unintended operations in future interactions, which differs from the immediate hijacking of prompt injection.79% similarUnverified持久化记忆是实现人工介入（Human-in-the-Loop）和断点续传的核心机制，Agent 暂停时状态被持久化，审批完成后可从中断处恢复72% similarVerifiedThe Reflexion framework enables Agents to reflect and summarize after failures, storing experiences in memory to avoid repeating mistakes, representing a further enhancement built on top of ReAct.70% similarUnverified推测执行过程中即使指令最终被回滚，其对缓存状态的影响仍会残留，攻击者可通过测量访存时延推断敏感数据67% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/745886API
curl https://kongchang.com/api/v1/knowledge/claims/745886MCP
get_claim(id=745886)