vLLM is an inference and serving engine for large language models. Prior to 0.32.0, ImageMediaIO passes attacker-controlled image bytes to Pillow through Image.open(BytesIO(data)) without restricting decoder selection, and the MIME label does not constrain the selected format. Image-capable request paths through parse_image() and fetch_image() can therefore dispatch an EPS payload to host components. A host with Ghostscript affected by CVE-2024-29510 can execute commands with the vLLM process's privileges, while environments using Pillow 12.0.0 through 12.2.0 can trigger CVE-2026-59203 and exhaust media-loading workers. Deployments without the applicable vulnerable dependency are not affected by that corresponding impact. This issue is fixed in version 0.32.0.
CVE-2026-113786 is an improper input validation vulnerability in the py-pdf pypdf library prior to version 6.21.1. An attacker can craft a PDF with image data labeled with the /DCTDecode filter that causes the Pillow library to invoke an unintended image parser when loading the image. This issue only affects environments where the optional Pillow dependency is installed. The vulnerability is fixed in pypdf version 6.21.1.
CVE-2026-96229 is a vulnerability in Microsoft .NET 10.0 runtime involving a permissive list of allowed inputs that could allow an unauthorized attacker to disclose information over a network. The flaw affects multiple .NET versions prior to their respective fixed releases. The vulnerability has a medium severity with a CVSS score of 6.5. An official fix is available and should be applied to mitigate the issue.
SAP UI5 does not perform sufficient input validation under certain configurations. An attacker with low privileges could submit specially crafted content that is later rendered in a victim's browser. This vulnerability has a low impact on confidentiality and integrity. There is no impact on availability.
A Security Policy Bypass vulnerability exists in Forcepoint Security Engine (NGFW).
This issue affects Forcepoint Security Engine (NGFW): from 7.1.0 through 7.1.13, from 7.3.0 through 7.3.1, 7.3.3, from 7.4.0 through 7.4.1, and 7.5.0.
CVE-2026-54694 is a critical improper input validation vulnerability in the NationalSecurityAgency's skills-service (SkillTree platform) prior to version 4.4.2. It involves unfiltered HTML input in user registration fields that leads to cross-site scripting (XSS) attacks when administrators view quiz runs. There are three escalating exploitation paths: basic XSS, remote script loading allowing arbitrary admin actions, and full cross-site request forgery (CSRF) token theft enabling API abuse. The vulnerability is patched in version 4.4.2.
CVE-2026-55581 is an OS command injection vulnerability in sonirico mcp-shell prior to version 0.6.0. The issue arises because the default Docker security.yaml allows /bin/bash in allowed_executables, but validation only checks the first token and does not block the shell command-mode flag -c. This allows an attacker to execute arbitrary commands as the mcpuser. The vulnerability is fixed in version 0.6.0.
The Apify MCP server enables AI agents to extract data from websites using ready-made scrapers, crawlers, and automation tools available on the Apify Store. Prior to 0.9.21, the fetch-apify-docs tool in src/tools/common/fetch_apify_docs.ts validates allowlisted documentation domains with String.startsWith() rather than URL hostname comparison, allowing attacker-controlled URLs such as `https://docs.apify.com.evil.com/` and `https://docs.apify.com@evil.com/` to pass the ALLOWED_DOC_DOMAINS check and return arbitrary fetched content to the LLM. This issue is fixed in version 0.9.21.
A vulnerability was found in nextlevelbuilder GoClaw 3.11.3. Affected by this issue is the function ExecApprovalManager.CheckCommand of the file internal/tools/exec_approval.go. The manipulation results in incomplete blacklist. The attack can be executed remotely. The exploit has been made public and could be used.
Glances is an open-source system cross-platform monitoring tool. Prior to 4.5.5, the Glances XML-RPC server (glances -s) introduced a configurable CORS origin list in version 4.5.3 as a mitigation for CVE-2026-33533. However, the implementation silently falls back to Access-Control-Allow-Origin: * whenever cors_origins contains more than one entry. An operator who configures an explicit two-entry allowlist (e.g. two internal dashboard origins) intending to restrict browser access instead receives the unrestricted wildcard. A malicious web page served from any origin can issue a CORS simple request to /RPC2 and read the full system monitoring dataset without the victim's knowledge. This vulnerability is fixed in 4.5.5.
Appsmith is a platform to build admin panels, internal tools, and dashboards. Prior to 2.1, Appsmith's bundled supervisord exposes an XML-RPC interface on port 9001, reachable from outside the container via a Caddy reverse-proxy route at /supervisor/* on the public ingress. Combined with the APPSMITH_SUPERVISOR_PASSWORD exposed via GET /api/v1/admin/env, any authenticated administrator can send arbitrary XML-RPC calls to supervisord and execute OS commands inside the Docker container via twiddler.addProgramToGroup. This vulnerability is fixed in 2.1.
Claude Code is an agentic coding tool. From 0.2.54 until 2.1.163, because the hostname huggingface.co was pre-approved as a bare hostname for the WebFetch tool, any path on that domain—including attacker-controlled model repositories—was auto-approved without a permission prompt or being subject to --allowedTools restrictions. An attacker able to inject untrusted content into a Claude Code context could direct it to issue WebFetch requests against attacker-controlled repository files (e.g. /resolve/main/config.json), which HuggingFace counts as downloads server-side, creating a covert out-of-band channel for encoding and exfiltrating data Claude can access such as files, environment variables, or command output. Reliably exploiting this required the ability to add untrusted content into a Claude Code context window. This vulnerability is fixed in 2.1.163.
A permissive list of allowed inputs in ASUS Armoury Crate allows a local administrator to perform arbitrary memory read/write operations or cause a system crash (BSOD) by bypassing the validation mechanism.Refer to the '
Security Update for Armoury Crate App ' section on the ASUS Security Advisory for more information.
CVE-2026-3490 is a critical vulnerability in picklescan versions before 1.0.4. The flaw allows attackers to bypass the blocklist by exploiting the permissive handling of pkgutil.resolve_name via indirect REDUCE calls. This enables remote code execution by invoking blocked functions such as os.system, builtins.exec, or subprocess.call.
HAX CMS helps manage microsite universe with PHP or NodeJs backends. Starting in version 9.0.1 and prior to version 26.0.0 of @haxtheweb/open-apis, multiple functions conduct substring-only matching to validate hostnames to which basic authorization should be sent. An attacker can append the matched substrings to an attacker-controlled endpoint and capture authentication. Version 26.0.0 fixes the issue.
OpenClaw versions before 2026.4.15 have an arbitrary file read vulnerability in the QMD backend memory_get function. This flaw allows users with access to the memory tool to bypass path restrictions and read any Markdown files within the workspace root, including files outside the intended canonical memory locations or indexed QMD result sets. The vulnerability is identified as CWE-183, relating to a permissive list of allowed inputs. The CVSS 4.0 base score is 2.3, indicating low severity.
OpenClaw versions before 2026.4.12 have an improper authorization vulnerability in helper-backed channels. The flaw causes empty resolved approver lists to be treated as explicit approval, allowing attackers with knowledge of an approval ID to bypass authorization and resolve pending approvals improperly.
NetBox versions 4.3.5 through 4.5.4 contain a remote code execution vulnerability in the RenderTemplateMixin.get_environment_params() method that allows authenticated users with exporttemplate or configtemplate permissions to execute arbitrary code by specifying malicious Python callables in the environment_params field. Attackers can bypass Jinja2 SandboxedEnvironment protections by setting the finalize parameter to any importable Python callable such as subprocess.getoutput, which is invoked on every rendered expression outside the sandbox's call interception mechanism, achieving remote code execution as the NetBox service user.
OpenClaw versions before 2026.3.22 have a vulnerability due to incomplete sanitization of host environment variables in specific configuration files. This flaw allows attackers to override package-manager environment settings, potentially redirecting package resolution or runtime bootstrap processes to attacker-controlled infrastructure, enabling execution of malicious content.
DOMPurify is a DOM-only cross-site scripting sanitizer for HTML, MathML, and SVG. Versions prior to 3.4.0 have an inconsistency between FORBID_TAGS and FORBID_ATTR handling when function-based ADD_TAGS is used. Commit c361baa added an early exit for FORBID_ATTR at line 1214. The same fix was not applied to FORBID_TAGS. At line 1118-1123, when EXTRA_ELEMENT_HANDLING.tagCheck returns true, the short-circuit evaluation skips the FORBID_TAGS check entirely. This allows forbidden elements to survive sanitization with their attributes intact. Version 3.4.0 patches the issue.