arXiv:2611.22075v1 Announce Type: new Abstract: Tool-using AI agents change files, databases, access rights and public information, yet permission to call a tool does not establish whether a mistake can be undone. We present Reversibility Classes, a declared five-level vocabular…
⏱ Độ trễ thật của tab này: trung bình 6.8 giờ sau khi CVE được công bố (n = 5441, sớm nhất 0.1 giờ). Số âm nghĩa là ta thấy trước ngày công bố.
Large language models are deployed as long-horizon agents that compose diverse tools across MCP services into executable workflows. Existing MCP environment-scaling methods are data-centric: they first synthesize a database and scenario, then derive tools and tasks from them. As …
arXiv:2610.27234v1 Announce Type: new Abstract: Enterprise adoption of LLMs in agentic workflows is gated by evaluations that span a wide arc of abilities: CRM operations, heterogeneous-document retrieval, personal data understanding, multi-turn planning, tool use, function call…
arXiv:2610.31974v1 Announce Type: new Abstract: Agentic AI systems using the Model Context Protocol (MCP) are vulnerable to indirect prompt injection through untrusted tool outputs. We demonstrate cross-modal trust laundering, in which instructions concealed in an email-signatur…
arXiv:2610.36108v1 Announce Type: new Abstract: LLM agents increasingly access external tools through the Model Context Protocol (MCP), but the natural-language metadata used to describe those tools creates a prompt-injection surface. Within this established tool-poisoning surfa…
arXiv:2610.41885v1 Announce Type: new Abstract: Large language model (LLM)-based agents interact with external tools through protocols such as the Model Context Protocol (MCP), producing tool-call traces that may expose behavioral signals about users. In this work, we systematic…
arXiv:2611.00729v1 Announce Type: new Abstract: Language-model agents commonly collapse two decisions that security systems separate: selecting an action and determining whether that action is authorized. When a model exercises standing authority, indirect prompt injection can c…
arXiv:2611.03662v1 Announce Type: new Abstract: The Model Context Protocol (MCP) connects large language model agents to external tools and services through official SDKs that implement message validation and session logic. Specification constraints omitted by these SDKs may pro…
arXiv:2611.06585v1 Announce Type: new Abstract: AI agents rely on third-party tools, such as Model Context Protocol (MCP) servers, whose publishers can be authenticated and whose execution can be restricted by runtime security policies. Yet authenticity does not guarantee benign…
arXiv:2611.07508v1 Announce Type: new Abstract: AI agents that work for an individual over time may also help that individual serve external clients. We define a personal external service agent as an agent authorized and continuously managed by a specific individual to accept cl…
arXiv:2611.07689v1 Announce Type: new Abstract: Large language model agents are increasingly deployed through the Model Context Protocol (MCP), which allows them to discover and invoke tools across multiple servers at runtime. Existing MCP benchmarks primarily measure task compl…
arXiv:2611.11104v1 Announce Type: new Abstract: As a core protocol for integrating large language models (LLMs) with external data sources and tools, the Model Context Protocol (MCP) has quickly become a key interface in emerging AI agent systems. As the MCP ecosystem expands, r…
Language agents increasingly act through external tools and interactive environments, making the interface between the model and its executable capabilities part of the decision process. Model Context Protocol (MCP) and command-line interfaces (CLI) can expose the same semantic o…
Paper này so sánh chi tiết chiến lược tích hợp công cụ CLI và MCP cho các hệ thống AI tự trị trên 14 tiêu chí, bao gồm đánh giá bảo mật và mô hình kinh tế, cung cấp cái nhìn thiết thực để developer lựa chọn phương án tích hợp tool.
arXiv:2610.33633v1 Announce Type: new Abstract: Agents select tools by copyable descriptions, which cannot establish which implementation runs. ProbeID compares canonicalized responses to schema-valid probes with a trusted admission reference and abstains on ambiguity, so a squa…
Bài paper này thực hiện benchmark thực tế đầu tiên về chi phí handshake (overhead thiết lập kết nối) trong các protocol giao tiếp agent hàng đầu như MCP, A2A và ACP, giúp các nhà phát triển AI agent đánh giá hiệu suất trao đổi dữ liệu ban đầu.
Bài báo arXiv giới thiệu Phronix, một learning paradigm mới cho AI agent, giúp agent tự học và giữ lại kiến thức từ các lần thực thi trước thay vì phải rediscover solutions.
arXiv:2610.33198v1 Announce Type: new Abstract: A neural classifier can fail completely inside a layered defense without any end-to-end number moving. We demonstrate this on BARBICAN, a fail-closed gateway for language-model tool mediation whose arbitration stage spent an entire…