26 vulnerabilities across 15 products scored HIGH or above on August 21, 2026.

  • 🔴 CRITICAL: 1
  • 🟠 HIGH: 25

🔴 [CRITICAL] spip/spip

1 CVE | CVSS 3.1: 9.8 | AAS 13.9

  • cpe:2.3:a:spip:spip:*:*:*:*:*:*:*:*

SPIP — Critical Remote Code Execution (Exploited in the Wild)

SPIP, the open-source content management and publishing system, is affected by one critical vulnerability (CVE-2026-77806, CVSS 9.8) that allows unauthenticated remote attackers to execute arbitrary code via crafted X-Spip-Filtre HTTP request headers due to improper handling in the template analysis function. This vulnerability is actively being exploited in the wild as of August 2026. All organizations running SPIP versions prior to 4.4.21 should treat this as an emergency and upgrade immediately by following the guidance in the vendor’s security advisory at blog.spip.net.

Vendor Advisory

🟠 [HIGH] libvnc/libvncserver

1 CVE | CVSS 3.1: 8.8 | AAS 11.4

  • cpe:2.3:a:libvnc:libvncserver:*:*:*:*:*:*:*:* (>= 0.9.12)

LibVNCServer — High-Severity Heap Out-of-Bounds Write via Malicious Server

LibVNCServer versions 0.9.12 through 0.9.15 are affected by one high-severity vulnerability (CVE-2026-50538, CVSS 8.8) in which a malicious or man-in-the-middle VNC server can trigger an out-of-bounds heap write on any connecting libvncclient with attacker-controlled length, contents, and offset, requiring no authentication and exploitable in a default build via a single FramebufferUpdate message. Any application or product embedding libvncclient for outbound VNC connections should be considered at risk. Teams should apply the vendor-provided fix immediately by updating to the patched commit available on the LibVNC GitHub repository.

Vendor Advisory

🟠 [HIGH] paperclipai/paperclip

1 CVE | CVSS 4.0: 9.4 | AAS 10.8

  • cpe:2.3:a:paperclipai:paperclip:*:*:*:*:*:*:*:*

Paperclip — Remote Command Execution via DNS Rebinding

Paperclip, an AI coding assistant by PaperclipAI, versions prior to 0.3.1 are affected by one high-severity vulnerability (CVE-2026-77087, CVSS 9.4) that enables remote command execution through DNS rebinding due to missing Host header validation in the default local_trusted mode. When a developer running Paperclip locally visits a malicious webpage, an attacker can exploit DNS rebinding to make authenticated API requests and execute arbitrary commands through the process adapter. Development teams using Paperclip should upgrade to version 0.3.1 or later immediately and review the vendor advisory on GitHub for additional mitigation guidance.

Vendor Advisory

🟠 [HIGH] nezhahq/nezha

1 CVE | CVSS 3.1: 9.9 | AAS 10.7

  • cpe:2.3:a:nezhahq:nezha:*:*:*:*:*:*:*:*

Nezha Monitoring — Stream Hijacking via Broken Access Control

Nezha Monitoring, a self-hostable server and website monitoring tool, versions 1.14.13 through 1.14.14 and 2.0.0 through 2.0.9 are affected by one high-severity vulnerability (CVE-2026-62283, CVSS 9.9) in which stream identifiers are not bound to their creating user, allowing any authenticated member who obtains a live stream UUID to hijack another user’s terminal or file transfer session. This broken access control in the WebSocket terminal and file endpoints could enable unauthorized command execution or data access on monitored servers. Administrators should update to the patched version referenced in the vendor’s GitHub commit immediately and audit access logs for any unauthorized stream connections.

Vendor Advisory

🟠 [HIGH] geotools/geotools

1 CVE | CVSS 3.1: 9.8 | AAS 10.6

  • cpe:2.3:a:geotools:geotools:*:*:*:*:*:*:*:*

GeoTools — SQL Injection in PostGIS DataStore OGC Filter Handling

GeoTools, a widely used open-source Java library for geospatial data processing, versions 30.5 through 34.x are affected by one high-severity SQL injection vulnerability (CVE-2026-76904, CVSS 9.8) in the PostGIS DataStore implementation, where the jsonArrayContains function writes user-supplied values into generated SQL without proper escaping on systems running PostGIS 12 or greater. Organizations using GeoTools with PostGIS backends for OGC-compliant geospatial services, including GeoServer deployments, should treat this as critical given the unauthenticated remote exploitation potential. Update immediately to GeoTools versions 33.6 or 34.5 and review the vendor’s GitHub commit for patch details.

Vendor Advisory

🟠 [HIGH] phalcon/cphalcon

1 CVE | CVSS 4.0: 9.2 | AAS 10.5

  • cpe:2.3:a:phalcon:cphalcon:*:*:*:*:*:*:*:*

Phalcon — PHP Code Injection via Volt Template Compiler

Phalcon, a high-performance full-stack PHP framework, versions 5.15.0 and earlier are affected by one high-severity vulnerability (CVE-2026-59989, CVSS 9.2) in which the Volt template compiler’s resolveFilter function inserts raw separator and array token values into generated PHP code without proper sanitization, allowing an attacker who can influence Volt template source to inject arbitrary PHP into compiled cache files and achieve remote code execution when the template is rendered. Teams running applications built on the Phalcon framework with user-controllable or untrusted template content are at immediate risk. Apply the fix referenced in the vendor’s pull request on GitHub and audit any environments where template sources may be influenced by external input.

Vendor Advisory

🟠 [HIGH] ggml-org/llama.cpp

1 CVE | CVSS 4.0: 9.2 | AAS 10.3

  • cpe:2.3:a:ggml-org:llama.cpp:*:*:*:*:*:*:*:*

llama.cpp — Unauthenticated Remote Code Execution via RPC Server Use-After-Free

llama.cpp versions prior to b8585 are affected by one high-severity vulnerability (CVE-2026-39909, CVSS 9.2) in which the RPC server’s GRAPH_RECOMPUTE handler contains a use-after-free condition that allows unauthenticated remote attackers to achieve arbitrary memory read and write access by storing a computation graph, freeing its referenced buffers, and reclaiming the freed memory with attacker-controlled content, leading to full remote code execution. Any organization exposing llama.cpp’s RPC server interface, particularly in networked or multi-node inference deployments, is at critical risk. Update to build b8585 or later immediately and ensure RPC endpoints are not exposed to untrusted networks.

Vendor Advisory

🟠 [HIGH] lxc/incus

5 CVEs | CVSS 3.1: 9.9 | AAS 10.2

  • cpe:2.3:a:lxc:incus:*:*:*:*:*:*:*:* (< 7.3.0)

Incus — Multiple Project Restriction and Container Escape Vulnerabilities

Incus, the LXC-based system container and virtual machine manager, versions prior to 7.3.0 are affected by five high-severity vulnerabilities (including CVE-2026-62941, CVSS 9.9) that collectively allow authenticated users to bypass project security restrictions during cross-project instance operations, enabling the injection of dangerous configuration keys such as security.privileged, raw.lxc, and raw.apparmor after authorization checks have passed, potentially leading to container escape and host compromise. Organizations using Incus in multi-tenant or project-isolated environments should treat this as critical, as the flaws undermine the core security boundary between projects. Upgrade to Incus 7.3.0 immediately and review the vendor’s security advisories on GitHub for details on all five issues.

Vendor Advisory

🟠 [HIGH] jsonata-js/jsonata

3 CVEs | CVSS 4.0: 9.3 | AAS 9.7

  • cpe:2.3:a:jsonata:jsonata:*:*:*:*:*:*:*:* (< 1.8.8)
  • cpe:2.3:a:jsonata:jsonata:*:*:*:*:*:*:*:* (>= 2.0.0, < 2.2.1)
  • cpe:2.3:a:jsonata-js:jsonata:*:*:*:*:*:*:*:*

JSONata — Prototype Pollution to Remote Code Execution via Crafted Expressions

JSONata, a widely used JSON query and transformation library for JavaScript, versions prior to 1.8.8 and 2.2.0 are affected by three high-severity vulnerabilities (including CVE-2026-77413, CVSS 9.3) in which missing prototype property checks in the lookup function allow crafted expressions to traverse inherited prototype members, ultimately reaching the child_process module to execute arbitrary commands with the privileges of the host Node.js process. Any application that evaluates user-supplied or untrusted JSONata expressions is at critical risk of full remote code execution. Update to JSONata 1.8.8 or 2.2.0 immediately and review the vendor’s security advisories on GitHub for details on all three issues.

Vendor Advisory

🟠 [HIGH] combodo/itop

3 CVEs | CVSS 3.1: 8.8 | AAS 9.6

  • cpe:2.3:a:combodo:itop:*:*:*:*:*:*:*:* (< 3.2.3)

Combodo iTop — Multiple Cross-Site Scripting and Web Security Vulnerabilities

Combodo iTop, a web-based IT service management platform, versions prior to 3.2.3 are affected by three high-severity vulnerabilities (including CVE-2026-33240, CVSS 8.8) that include reflected cross-site scripting in the foreign key search criteria API, which could allow attackers to execute malicious scripts in the context of authenticated ITSM user sessions. Organizations relying on iTop for IT service management and asset tracking should prioritize remediation, as exploitation could lead to session hijacking, privilege escalation, or unauthorized access to sensitive operational data. Upgrade to iTop version 3.2.3 immediately and review the vendor’s GitHub commits for details on all three fixes.

Vendor Advisory

🟠 [HIGH] 101gen/automation_web_platform_–_notifications_and_otp_for_woocommerce,_advanced_country_code

1 CVE | CVSS 3.1: 9.8 | AAS 9.6

  • cpe:2.3:a:101gen:automation_web_platform_notifications_and_otp_for_woocommerce_advanced_country_code:*:*:*:*:*:*:*:*

101gen Automation Web Platform — Authentication Bypass via Exposed Magic Login Token

The Automation Web Platform – Notifications and OTP for WooCommerce plugin for WordPress, versions up to and including 4.8.6, is affected by one high-severity authentication bypass vulnerability (CVE-2026-77264, CVSS 9.8) in which the handle_email_otp_return function returns the secret magic login token directly in the API response to a publicly accessible OTP request instead of delivering it exclusively via email, allowing unauthenticated attackers to log in as any user including administrators. Any WordPress site running this plugin with OTP login functionality enabled is at immediate risk of full site takeover. Site administrators should update the plugin beyond version 4.8.6 immediately or deactivate it until a patch is applied, and audit for any signs of unauthorized access.

Vendor Advisory

🟠 [HIGH] tyrrrz/discordchatexporter

1 CVE | CVSS 3.1: 8.2 | AAS 9.5

  • cpe:2.3:a:tyrrrz:discordchatexporter:*:*:*:*:*:*:*:* (< 2.47.2)

DiscordChatExporter — Stored Cross-Site Scripting in HTML Chat Exports

DiscordChatExporter, a tool for saving Discord chat logs to file, versions prior to 2.47.2 are affected by one high-severity vulnerability (CVE-2026-54682, CVSS 8.2) in which HTML exports generated with markdown formatting disabled render attacker-controlled message content, embed titles, descriptions, and field values without HTML entity encoding, enabling stored cross-site scripting that executes when anyone opens the exported file in a browser. Any user or organization exporting Discord channels that contain untrusted user-generated content is at risk of arbitrary script execution in the viewer’s browser context. Update to DiscordChatExporter version 2.47.2 or later, and treat any previously exported HTML files from untrusted channels as potentially compromised.

Vendor Advisory

🟠 [HIGH] omnigent-ai/omnigent

3 CVEs | CVSS 3.1: 9.0 | AAS 9.3

  • cpe:2.3:a:omnigent-ai:omnigent:*:*:*:*:*:*:*:* (< 0.3.0)

Omnigent — Shared Agent Poisoning and Privilege Escalation

Omnigent, an open-source AI agent framework for orchestrating coding agents, versions prior to 0.3.0 are affected by three high-severity vulnerabilities (including CVE-2026-62674, CVSS 9.0) in which insufficient authorization checks on shared and template agent bundles allow an authenticated user with session edit access to replace a shared agent’s configuration, inject a malicious MCP server, and compromise all subsequent sessions that reference that agent. Organizations using Omnigent in multi-user or shared environments face a supply-chain-style risk where a single compromised agent bundle can propagate arbitrary code execution across users. Update to Omnigent 0.3.0 or later immediately and audit any shared agent configurations for unauthorized modifications.

Vendor Advisory

🟠 [HIGH] zanllp/infinite-image-browsing

2 CVEs | CVSS 4.0: 8.7 | AAS 9.1

  • cpe:2.3:a:zanllp:infinite-image-browsing:*:*:*:*:*:*:*:*

Infinite Image Browsing — Path Traversal via Directory Confinement Bypass

Infinite Image Browsing, a web-based image browsing extension commonly used with Stable Diffusion interfaces, is affected by two high-severity vulnerabilities (including CVE-2026-77814, CVSS 8.7) in which the path trust validation in the API uses a flawed startswith comparison without appending a path separator, allowing attackers to access files in sibling directories whose names share a common prefix with an allowed path, effectively bypassing the intended directory confinement to disclose sensitive files. Anyone running this tool with network exposure, particularly in shared or multi-user AI image generation environments, is at risk of unauthorized file disclosure beyond configured boundaries. Users should monitor the vendor’s GitHub repository for a patched release and restrict network access to the application to trusted users only until a fix is available.

Vendor Advisory

🟠 [HIGH] getkin/kin-openapi

1 CVE | CVSS 4.0: 8.7 | AAS 9.1

  • cpe:2.3:a:getkin:kin-openapi:*:*:*:*:*:*:*:* (>= 0.124.0, < 0.142.0)

kin-openapi — Denial of Service via Unbounded Memory Allocation in Query Parameter Parsing

kin-openapi, a widely used Go library for handling OpenAPI specifications, versions 0.124.0 through 0.141.x are affected by one high-severity vulnerability (CVE-2026-77354, CVSS 8.7) in which an unauthenticated attacker can trigger unbounded memory allocation by sending crafted deepObject query parameters with sparse indexes, as the request decoder converts these into dense slices before schema validation occurs, bypassing any maxItems constraints. Any API service using kin-openapi’s request validation middleware is vulnerable to remote denial of service through a single malicious request. Update to version 0.142.0 or later immediately and review the vendor’s security advisory on GitHub for further details.

Vendor Advisory