intentshield

IntentShield Don't filter what your AI says. Filter what it's about to do Pre-execution intent verification for AI agents. Why This Exists AI agents have tool access. They can execute shell commands, write files, browse URLs, send emails, and call APIs. Every one of those actions is a potential...

oversight

Oversight v0.5.8 Oversight は、シェルスクリプトやコマンドがシステムに影響を及ぼす前に監査するために設計された、ターミナルベースのセキュリティインテリジェンスツールです。「盲目的な信頼」と「手動監査」の間を埋めるもので、ローカルファイルと Raw URL を介したリモートスクリプトの両方を高速かつ対話的に分析します。 Oversight Live AUR Update Screenshot 1...

sqlsure

sqlsure AI 写 SQL。sqlsure 确保它是正确的。 查询可以完全有效，无错误运行，并返回一个数字，但该数字可能是悄悄错误的——因为连接导致收入重复计算，平均值被求和，或者患者标识符被暴露。数据库不捕捉这些。代码检查工具不捕捉这些。LLM 审查自己的 SQL 也不捕捉这些。 sqlsure 可以——确定性、在 0.1 毫秒内、在查询运行之前做到这一点。 证明，而非承诺： 我们在每个文本到 SQL 模型评分的两个基准测试的金标准答案上运行了 sqlsure。2,568 个专家编写的查询，45 个标记，零误报 ——包括一个 BIRD 开发金标准答案，该答案因 sqlsure 针...

authproof-sdk

AuthProof SDK AuthProof 是一种面向代理型 AI 的加密授权协议。该领域的大多数协议都基于操作员定义的策略执行——赋予操作员在事后扩展或重新解释用户原始意图的权力。AuthProof 基于不同的信任模型构建：用户自己的私钥签署控制执行的授权对象，并且在授权时和执行前立即验证实时模型状态。用户签署的权限与实时模型状态门控的结合是具体的主张——而非广泛的执行叙事。 不同之处在于： 用户是签署权威。 每个竞争协议（AIP、AITH、OAP、SAGA、AgentSpec）都基于操作员定义的策略执行。在 AuthProof...

CVE-2026-72774

n8n before 1.123.67, 2.31.5, and 2.32.1 contains a credential authorization bypass in the HTTP Request node. An authenticated member with edit access to a shared workflow can reference another user's credential while specifying the credential type via an expression. Because the pre-execution...

Temporal UI State Inconsistency in Desktop GUI Agents: Formalizing and Defending against TOCTOU Attacks on Computer-Use Agents

GUI agents that control desktop computers via screenshot-and-click loops introduce a new class of vulnerability: the observation-to-action gap mean 6.51 s on real OSWorld workloads creates a Time-Of-Check, Time-Of-Use TOCTOU window during which an unprivileged attacker can manipulate the UI state...

From Component Manipulation to System Compromise: Understanding and Detecting Malicious MCP Servers

The model context protocol MCP standardizes how LLMs connect to external tools and data sources, enabling faster integration but introducing new attack vectors. Despite the growing adoption of MCP, existing MCP security studies classify attacks by their observable effects, obscuring how attacks...

CVE-2026-32921

OpenClaw CVE-2026-32921 affects OpenClaw prior to version 2026.3.8. The vulnerability is an approval bypass in system.run where mutable script operands are not bound across the approval and execution phases. An attacker who gains approval for script execution can modify the approved script file b...

EUVD-2026-15943

n8n is Vulnerable to Credential Theft via Name-Based Resolution and Permission Checker Bypass in Community Edition...

Microsoft Windows 11 Build 26200 File Explorer Auditor

This Metasploit module provides a defensive pre-execution assessment for the Windows vulnerability where File Explorer fails to properly restrict access to sensitive system locations...

AEGIS: No Tool Call Left Unchecked -- a Pre-Execution Firewall and Audit Layer for AI Agents

AI agents increasingly act through external tools: they query databases, execute shell commands, read and write files, and send network requests. Yet in most current agent stacks, model-generated tool calls are handed to the execution layer with no framework-agnostic control point in between...

Atomicity for Agents: Exposing, Exploiting, and Mitigating TOCTOU Vulnerabilities in Browser-Use Agents

Browser-use agents are widely used for everyday tasks. They enable automated interaction with web pages through structured DOM based interfaces or vision language models operating on page screenshots. However, web pages often change between planning and execution, causing agents to execute action...