1
🔥 I’m Done Building ATLOCK Just to “Make It Work.”
ATLOCK开发者放弃堆功能，转向攻击者视角验证安全假设的硬核思考。
I want to build something actually worth using. When I started ATLOCK, the goal was simple: Build a Windows security application. But somewhere along …
2
Bounding the Blast Radius: A Survey of Prompt-Injection Defenses for LLM Agents
别指望魔法防御，提示注入至今无人完美解决，这篇调查把现有方案的死角都翻了个底朝天。
Article URL: https://fabraix.com/blog/nobody-has-solved-prompt-injection Comments URL: https://news.ycombinator.com/item?id=48802392 Points: 1 # Comme…
3
Exploiting LLM Agent Supply Chains via Payload-Less Skills
揭示LLM Agent供应链中第三方技能的无payload攻击风险，为系统安全评估敲响警钟。
Article URL: https://arxiv.org/abs/2605.14460 Comments URL: https://news.ycombinator.com/item?id=48789488 Points: 2 # Comments: 0
4
SMSR: Certified Defence Against Runtime Memory Poisoning in Persistent LLM Agent Systems
首个认证防御方案，针对LLM代理系统运行时内存投毒攻击，保障持久化RAG安全。
arXiv:2606.12703v1 Announce Type: cross Abstract: Retrieval-augmented generation (RAG) agents increasingly run with persistent memory that accumulates…
5
The Meta hack shows there’s more to AI security than Mythos
Meta黑客事件揭示AI安全不止于技术神话，自动化工作流正成攻击新靶心。
On June 5, 404 Media reported that attackers had been using Meta’s AI customer support agent to steal Instagram accounts. Their approach was simple: T…