Unverified50% confidenceFactExact time
When an agent fails to execute a dangerous operation in the face of a Prompt Injection attack, this may stem from two paths: the security policy catching the malicious path, or the model avoiding the tool or task for unrelated reasons.
1
Sources
50%
Confidence
Long-term
Relevance
9/11/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedIf the agent completes the task normally in the benign version but refuses the dangerous call in the malicious version, it is stronger evidence that the security policy is working.78% similarUnverifiedAgent链式调用多个技能时，若某个技能被恶意构造或存在漏洞，可能引发提示注入攻击（Prompt Injection）76% similarUnverified成熟的 Agent 框架通常通过沙箱隔离、权限最小化和执行前确认等机制管控脚本执行的安全风险74% similarUnverifiedThe `/goal` command's dual-agent architecture separates 'execution' and 'verification' into two independent agents to avoid blind spots that occur when a single model evaluates its own work72% similarUnverifiedAgents may execute destructive operations far beyond what was intended due to hallucination or prompt injection attacks.72% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/901656API
curl https://kongchang.com/api/v1/knowledge/claims/901656MCP
get_claim(id=901656)