CVE-2026-87798 LXD client recursive file pull allows directory escape via malicious VM agent

Improper link resolution in the recursive file pull feature of the LXD CLI client in Canonical LXD versions 4.0.2 up to 6.9 fixed in 4.0.14, 5.0.10 and 5.21.8 on Linux allows an attacker with root access inside a virtual machine to write attacker-controlled files or directory trees to arbitrary...

Carbonato Botnet Compromises Docker Hosts to Deploy Telegram-Controlled Hermes AI Agent

Cybersecurity researchers have disclosed details of a new botnet malware called Carbonato that's targeting exposed Docker daemons to deploy an open-source artificial intelligence AI agent framework called Hermes Agent. "The implant installs the framework unchanged, then overwrites its SOUL.md...

CVE-2026-84651

A flaw was found in Jenkins. The system's application programming interface API and command-line interface CLI for managing agent configurations fail to properly validate the target agent. This vulnerability allows an attacker, who already has permissions to configure one agent, to overwrite the...

CVE-2026-84648

A flaw was found in Jenkins. The system log viewer does not properly escape log record metadata, such as the source, level, and timestamp. This vulnerability allows attackers who control agent processes to inject malicious scripts. Successful exploitation can lead to a stored cross-site scripting...

Security Bulletin: IBM Instana Observability is affected by multiple vulnerabilities within Instana Agent container image

Summary Multiple vulnerabilities were addressed in IBM Observability with Instana within Instana Agent container image build 1.0.325 Vulnerability Details CVEID:CVE-2026-5450 DESCRIPTION: Calling the scanf family of functions with a %mc malloc'd character match in the GNU C Library version 2.7 to...

AI-Based Vulnerability Assessment Capability and Cyber Attack Graph Analysis

Cyber threats targeting mission-critical infrastructure are becoming more sophisticated while the barrier to launching attacks continues to fall. Traditional point solutions like antivirus and firewalls are reactive and fail to address the combinatorial complexity of modern attack surfaces. This...

CounterSteer: Suppressing Indirect Prompt Injection with Activation Steering

Indirect prompt injection makes an LLM agent treat untrusted retrieved text as instructions. We present CounterSteer, an inference-time defense that suppresses this behavior inside the model. Per model, a five-step recipe fits a residual-stream direction from paired episodes differing only in...

DGF-Bench: A Benchmark for Simulating and Auditing Deception against Multi-Agent Governance Boards

Tool-using language-model agents can review enterprise projects as governance boards do: they read the evidence, apply written rules and decide whether the project may proceed. Part of that evidence comes from suppliers and project members with a stake in the decision. DGF-Bench is a benchmark in...

Tracekit: Tamper-Evident Intent-Reasoning-Action Auditing for Autonomous Coding Agents

Autonomous coding agents read untrusted files, run shell commands and spawn sub-agents with little supervision, yet their record is usually an editable log. We present Tracekit, an open-source, dependency-free system that captures three channels for every agent session: what the human asked inten...

Xalgorix Autonomous AI Pentesting Agent 4.6.113

Most scanners detect. Xalgorix proves. An autonomous LLM agent works a full pentest methodology, then an independent verifier re-exploits every finding before it's reported - so you get proof, not a pile of maybes to triage. Self-hosted, private, and bring-your-own-LLM. Built in Go + TypeScript...

CVE-2026-87798: Improper Link Resolution Before File Access

Improper link resolution in the recursive file pull feature of the LXD CLI client in Canonical LXD versions 4.0.2 up to 6.9 fixed in 4.0.14, 5.0.10 and 5.21.8 on Linux allows an attacker with root access inside a virtual machine to write attacker-controlled files or directory trees to arbitrary...

PT-2026-99894

Name of the Vulnerable Software and Affected Versions Canonical LXD versions 4.0.2 through 4.0.13 Canonical LXD versions 5.0.0 through 5.0.9 Canonical LXD versions 5.1.0 through 5.21.7 Canonical LXD versions 6.0.0 through 6.9 Description Improper link resolution in the recursive file pull feature...

CVE-2026-88808: Execution with Unnecessary Privileges

A vulnerability has been identified within Rancher Manager where the Fleet agent wrote resources to downstream clusters using its own cluster-admin credentials instead of the ServiceAccount pinned to the deployment. It affects multi-tenancy environments where different tenants share the same...

Important: amazon-ssm-agent

Issue Overview: Handshakes which used Encrypted Client Hello could be de-anonymized by a passive network observer due to a disclosure of pre-shared key identities in the unencrypted client hello. CVE-2026-42505 go-git is an extensible git implementation library written in pure Go. Prior to 5.19.2...

JEV As a Judge for Agent Trace Security: An Empirical Comparison with Generative LLM Judges

Security evaluation of tool-using agents requires judging actions in context, yet generative judges add latency, explanation overhead, and output-validation failures. We study whether JEV, a typed decision model, offers a useful alternative for retrospective trace classification. We evaluate JEV...

Exploit for CVE-2026-71963

CVE-2026-71963 — Hermes Agent RCE via a repository-delivered...

CVE-2026-101086 Nezha Dashboard before 2.3.5 Task Type Validation Bypass

Nezha Dashboard versions before 2.3.5 fail to restrict service monitor task types to supported probe types, allowing authenticated users with nezha:service:write scope to submit privileged task types through the service API. Attackers can deliver command execution or Agent configuration tasks to...

CVE-2026-101086 Nezha Dashboard before 2.3.5 Task Type Validation Bypass

Nezha Dashboard versions before 2.3.5 fail to restrict service monitor task types to supported probe types, allowing authenticated users with nezha:service:write scope to submit privileged task types through the service API. Attackers can deliver command execution or Agent configuration tasks to...

ActBench

ActBench ActBench is a self-evolving benchmark of behavioral safety in cowork agents. It defines behavioral safety as whether an agent's execution remains within the permissions and state changes required by a benign task, and evaluates realized behavioral risk from execution trajectories rather...

log4shell-cve-lab

Лаборатория Log4Shell CVE-2021-44228 Этот репозиторий содержит небольшой уязвимый HTTP-сервис на Log4j 2.14.1 для CVE-2021-44228. Sentinel воспроизвёл этот сценарий 2026-08-27 как лабораторию в песочнице, созданную с нуля. Важные свойства этого запуска сохранены здесь: контролируемый атакующим...