AgentDoG 1.5: A Lightweight and Scalable Alignment Framework for AI Agent Safety and Security

Modern open-world agents such as OpenClaw exhibit powerful cross-environment execution capabilities yet introduce broad new safety risk sources. Meanwhile, advanced frontier AI models drastically lower attack barriers, rendering current agent alignment frameworks inadequate for real-world...

Technical Report: Exploring the Emerging Threats of the Agent Skill Ecosystem

We analyzed 3,984 AI agent skills from major marketplaces and found 76 confirmed malicious payloads, including credential theft, backdoor installation, and data exfiltration. 13.4% of all skills contain at least one critical-level security issue and at least 8 manually confirmed malicious skills...

CVE-2026-4093

A flaw was found in the Drupal 7 Term Reference Tree module. This vulnerability, a type of stored Cross-Site Scripting XSS, allows an authenticated attacker with permissions to edit or create taxonomy terms to inject malicious scripts. These scripts can execute when a user views a form containing...

EUVD-2026-31377

In the Drupal 7 Term Reference Tree module, two stored XSS vectors exist in the widget/formatter rendering pipeline. Vector A token display templates: When the Token module is enabled and token display templates are configured, attacker-controlled token output e.g., term description is rendered...

CVE-2026-4093

In the Drupal 7 Term Reference Tree module, two stored XSS vectors exist in the widget/formatter rendering pipeline. Vector A token display templates: When the Token module is enabled and token display templates are configured, attacker-controlled token output e.g., term description is rendered...

CVE-2026-4093 Stored XSS in Drupal 7 Term Reference Tree module (token display templates and term labels)

In the Drupal 7 Term Reference Tree module, two stored XSS vectors exist in the widget/formatter rendering pipeline. Vector A token display templates: When the Token module is enabled and token display templates are configured, attacker-controlled token output e.g., term description is rendered...

CVE-2026-4093

CVE-2026-4093 is a stored XSS in the Drupal 7 Term Reference Tree module affecting versions up to and including 7.x-1.11. Two vectors are described: Vector A (token display templates): attacker-controlled token output (e.g., term description) is rendered without proper sanitization when the Token...

Refusal Evaluation in Coding LLMs and Code Agents: A Systematic Review of Thirteen Malicious-Code Prompt Corpora (2023-2025)

The evaluation of large language model refusal on malicious-coding tasks now spans at least thirteen publicly released prompt corpora AdvBench, the CyberSecEval family, RMCBench, RedCode, MCGMark, JailbreakBench, CySecBench, MalwareBench, CIRCLE, MOCHA, ASTRA, Scam2Prompt / Innoc2Scam-bench, and...

Behavioral Integrity Verification for AI Agent Skills

Agent skills extend LLM agents with privileged third-party capabilities such as filesystem access, credentials, network calls, and shell execution. Existing safety work catches malicious prompts and risky runtime actions, but the skill artifact itself goes unverified. We formalize this as the...

Iterative Audit Convergence in LLM-Managed Multi-Agent Systems: A Case Study in Prompt Engineering Quality Assurance

Prompt specifications for multi-agent large language model LLM systems carry data contracts and integration logic across many interdependent files but are rarely subjected to structured-inspection rigor. This paper reports a single-system empirical case study of iterative, agent-driven auditing...

Do Skill Descriptions Tell the Truth? Detecting Undisclosed Security Behaviors in Code-Backed LLM Skills

Programmatic skills in LLM ecosystems consist of a natural-language description and executable implementation files. Users and LLMs rely on the description to understand the skill's scope. However, the implementation may perform security-relevant operations, such as credential access, network...

CVE-2026-42842

CVE-2026-42842: Stored XSS in Grav Form plugin (select field) where taxonomy values render via Twig |raw in admin pages, enabling an editor-level user to inject JavaScript that runs in admins’ browsers when viewing/editing pages. Affects Grav CMS Form plugin’s select.html.twig handling and global...

CVE-2026-42842 grav-plugin-form: XSS via Taxonomy Field Values in Admin Panel

The form plugin for Grav adds the ability to create and use forms. Prior to 9.1.0, a Stored Cross-Site Scripting XSS vulnerability exists in the Grav CMS Form plugin's select field template. Taxonomy tag and category values are rendered with the Twig |raw filter in the admin panel, bypassing the...

CVE-2026-42842

The form plugin for Grav adds the ability to create and use forms. Prior to 9.1.0, a Stored Cross-Site Scripting XSS vulnerability exists in the Grav CMS Form plugin's select field template. Taxonomy tag and category values are rendered with the Twig |raw filter in the admin panel, bypassing the...

CVE-2026-42842 grav-plugin-form: XSS via Taxonomy Field Values in Admin Panel

The form plugin for Grav adds the ability to create and use forms. Prior to 9.1.0, a Stored Cross-Site Scripting XSS vulnerability exists in the Grav CMS Form plugin's select field template. Taxonomy tag and category values are rendered with the Twig |raw filter in the admin panel, bypassing the...

Security Risks in Tool-Enabled AI Agents: A Systematic Analysis of Privileged Execution Environments

Tool-enabled AI agents are increasingly deployed in cloud-hosted environments and offered as services, where they perform side-effecting operations through privileged tools within execution environments. While such agents enable powerful automation, the security implications of hosting autonomous...

MonitoringBench: Semi-Automated Red-Teaming for Agent Monitoring

We introduce a red-teaming methodology that exposes harder-to-catch attacks for coding-agent monitors, suggesting that current practices may under-elicit attacks and overstate monitor performance. We identify three challenges with current red-teaming. First, mode collapse in attack generation,...

When the Ruler Is Broken: Parsing-Induced Suppression in LLM-Based Security Log Evaluation

LLM-based SOC log classifiers are commonly evaluated using regular-expression pipelines that extract structured fields from free-form model output. We demonstrate that this practice introduces a class of silent, systematic evaluation errors, which we term parsing-induced suppression that can caus...

GHSA-C2Q3-P4JR-C55F Grav Vulnerable to XSS via Taxonomy Field Values in Admin Panel

Summary A Stored Cross-Site Scripting XSS vulnerability exists in the Grav CMS Form plugin's select field template. Taxonomy tag and category values are rendered with the Twig |raw filter in the admin panel, bypassing the global autoescape protection. An editor-level user can inject arbitrary...

Grav Vulnerable to XSS via Taxonomy Field Values in Admin Panel

Summary A Stored Cross-Site Scripting XSS vulnerability exists in the Grav CMS Form plugin's select field template. Taxonomy tag and category values are rendered with the Twig |raw filter in the admin panel, bypassing the global autoescape protection. An editor-level user can inject arbitrary...