Unverified85% confidenceFactTime unknown
DeadEnd-CLI采用Agentic（代理式）架构，AI能够根据目标环境反馈动态调整攻击策略，形成推理-行动循环（ReAct Loop）
1
Sources
85%
Confidence
Medium-term (~90 days)
Relevance
6/1/2026
First Seen
Valid until: 8/30/2026
Sources
DeadEnd-CLI：开源AI渗透测试工具黑盒基准达81%通过率
githubxoxruns
Related Entities
Related Claims
Unverified越狱手法可分为提示注入、对抗性前缀攻击、多步骤语境构建以及自动化越狱67% similarUnverified这类系统多采用人在回路或人在回路之上的设计，由人类操作员做出攻击决策并选定目标，AI仅负责末段追踪与撞击66% similarUnverified智能体系统采用感知-规划-执行的闭环架构，通过ReAct、Chain-of-Thought等推理框架将复杂任务分解为子任务序列65% similarUnverified该决策树方案在Skill执行失败时会自动切换到备用CLI工具重新执行64% similarUnverifiedReAct架构的核心循环为Thought→Action→Observation，攻击者可以在Observation阶段注入恶意内容64% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/29525API
curl https://kongchang.com/api/v1/knowledge/claims/29525MCP
get_claim(id=29525)