MaxKB is an open-source AI assistant for enterprise. Prior to version 2.10.6-lts, the ToolExecutor LD_PRELOAD sandbox hooks execve, execvpe, and execveat to prevent subprocess creation but does not hook fexecve. An authenticated attacker able to execute tool code can call fexecve to start a process outside the sandbox's intended subprocess policy. This issue is fixed in version 2.10.6-lts.
The Hide My WP Ghost WordPress plugin before 7.0.11 does not properly validate a loopback security-check request before disabling its login and URL hiding protection, dropping that protection precisely when the request's verification value is missing or incorrect, which any visitor can arrange, allowing unauthenticated attackers to re-expose the concealed WordPress login page location.
The Hide My WP Ghost WordPress plugin before 7.0.11 does not verify that a request is a genuine WooCommerce request before disabling its firewall, threat-detection and login/URL-hiding protections, treating the mere presence of an attacker-suppliable request parameter as sufficient, which allows unauthenticated attackers to disable those protections and re-expose the concealed login and admin URLs on any request.
mport is the MidnightBSD Package Manager. Prior to 2.7.8, the audit command in mport/mport.c computed option-adjusted local_argv and local_argc values but passed the original argument entry to audit_package(). When an operator or automation used an option such as -r before a package name, stale optind state and the unadjusted argument could cause mport to audit the option token instead of the requested package, producing a false-negative or useless result that could leave a vulnerable package unidentified. The corrected parsing resets optind and optreset before using the adjusted local arguments. This issue is fixed in version 2.7.8.
CVE-2026-76825 is a vulnerability in zopefoundation's RestrictedPython prior to version 8.4 that allows sandbox escape. The issue arises when restricted code can access the standard library string module's Formatter class or instances, which internally perform attribute and item traversal without RestrictedPython's safer_getattr protections. This enables restricted code to reach sensitive objects such as function globals, builtins, file access, or code execution primitives, compromising confidentiality, integrity, and availability. The vulnerability is fixed in version 8.4.
Zope AccessControl provides a general security framework for use in Zope. Prior to 7.4, applications that allow untrusted users to create and execute AccessControl-controlled Python code do not safely guard str.format and str.format_map when those methods are reached through a str subclass. In both ImplPython.py and cAccessControl.c, Python formatting can recursively access attributes and subscriptions using unrestricted getattr and getitem behavior instead of the policy-restricted getattr and getitem operations. A controlled format string can therefore disclose objects reachable from values available to the formatting operation. This issue is fixed in version 7.4.
MCP Context Forge is an AI gateway, registry, and proxy for MCP, A2A, REST, and gRPC APIs. Prior to 1.0.2, the python_sandbox_server in mcp-servers/python/python_sandbox_server/src/python_sandbox_server/server_fastmcp.py exposes raw getattr through safe_builtins, omits a required _getattr_ guard, and relies on validate_code checks for literal dangerous dunder strings. An attacker can construct dunder names at runtime, traverse the Python class hierarchy, reach subprocess.Popen, and execute OS commands with the server process privileges through the execute_code MCP tool. The HTTP/SSE transport can expose this tool without authentication, while stdio-only deployments have reduced network reachability. The issue affects the python_sandbox_server subproject and does not directly affect the core Context Forge gateway or proxy components. This issue is fixed in version 1.0.2.
CVE-2026-57133 is an OS command injection vulnerability in MervinPraison's PraisonAI multi-agent teams system. Versions from 1.5.1 up to but not including 1.7.2 are affected. The vulnerability arises because the shell() helper only validates the first whitespace-delimited token against a safe command list but then executes the entire original string, allowing an attacker to append additional commands. This can lead to arbitrary command execution with the privileges of the PraisonAI process. The issue is fixed in version 1.7.2.
PraisonAI is a multi-agent teams system. From 1.2.3 until 1.7.2, SandboxExecutor network-isolated mode in src/praisonai-ts/src/cli/features/sandbox-executor.ts uses buildEnv() only to inject invalid http_proxy and https_proxy environment variables and does not establish an operating-system network boundary. Programs that ignore those proxy variables can open sockets directly, allowing supposedly isolated commands to reach localhost, internal services, cloud metadata, or external hosts and potentially exfiltrate data. An initial remediation was released in version 1.7.2.
PraisonAI is a multi-agent teams system. From 1.4.0 until 1.7.2, codeMode in src/praisonai-ts/src/tools/builtins/code-mode.ts executes untrusted JavaScript with new Function() inside with(sandbox) and relies on a small source-code blocklist plus shadowed process and require properties. Code can use ({}).constructor.constructor to recover the real Function constructor, obtain process and process.mainModule.require, and reach host filesystem and subprocess APIs despite the advertised sandbox. Attackers who control codeMode input can read secrets, modify files, execute commands, or exhaust the host process. This issue is fixed in version 1.7.2.
PraisonAI is a multi-agent teams system. From 1.2.3 until 1.7.2, CommandValidator in src/praisonai-ts/src/cli/features/sandbox-executor.ts validates only the first whitespace-delimited executable against allowedCommands, then SandboxExecutor passes the complete command string to sh -c. A command beginning with an allowed executable can append a non-allowlisted command through shell metacharacters, causing arbitrary commands to run with the PraisonAI process privileges. This issue is fixed in version 1.7.2.
PraisonAI is a multi-agent teams system. From 1.4.0 until 1.7.2, createAgentLoop() in src/praisonai-ts/src/ai/agent-loop.ts passes executable tools to generateText() before invoking the onToolCall approval callback. Because the wrapped AI SDK executes tool handlers during generation, a callback that returns false records tool_rejected only after the denied tool has already produced side effects and populated toolResults. Applications using onToolCall as a human or policy approval boundary can therefore execute rejected file, command, API, or data-modifying operations. This issue is fixed in version 1.7.2.
OpenAM versions prior to 16.1.1 contain a protection mechanism failure in the GroovySandboxValueFilter that allows authenticated server-side script authors to escape the scripting sandbox. This vulnerability enables users with sub-realm RealmAdmin privileges who can create or edit scripts to execute operating system commands as the OpenAM server account, compromising the JVM and all realms it serves. The issue is resolved in OpenAM version 16.1.1.
PraisonAI is a multi-agent teams system. Prior to praisonaiagents 1.6.59, execute_code sandbox mode permits runtime assembly of blocklisted dunder names and allows str.format or str.format_map to resolve dotted fields through C-level attribute access that bypasses _safe_getattr. This exposes class, qualified-name, base-class, globals, and object-dictionary attributes to prompt-influenced code when approval is automatically granted, producing a high-impact read primitive without establishing a complete in-process execution chain. This issue is fixed in praisonaiagents 1.6.59.
IBM Langflow OSS 1.0.0 through 1.11.5 An attacker who could submit custom component source code could bypass the static security scanner by crafting an annotated class-body assignment that resolved to a dangerous callable through alias tracking; the resolved value was never checked against the dangerous callable blocklist due to the logic error. If the crafted component reached the runtime execution path, the attacker could cause arbitrary operating system commands to execute on the server in-process, with the privileges of the running service.
CVE-2026-45770 affects Suricata versions 8.0.0 up to but not including 8.0.5. A vulnerability exists where a Lua rule registering too many flow variables can corrupt the Lua detection state and bypass Suricata's restricted Lua sandbox. This may also cause Suricata to crash. The issue requires an affected Lua script or rule to be loaded. Version 8.0.5 includes a fix, and disabling 'security.lua.allow-rules' is a recommended workaround if Lua rules are not needed.
CVE-2026-0306 is a medium-severity vulnerability in Palo Alto Networks Prisma Access Agent affecting versions prior to 26.2. It allows a local user to bypass Data Loss Prevention (DLP) policy enforcement controls, potentially enabling exfiltration of sensitive data. This issue does not affect the Prisma Access Agent on macOS, Linux, iOS, Android, or Chrome OS. A fix is available in version 26.2.
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.
Dell SCG 5.0 Appliance versions prior to 5.36.00.16 and Dell SCG 5.0 Application versions prior to 5.36.00.00, contains an Improper Neutralization of Alternate XSS Syntax vulnerability. An unauthenticated attacker with remote access could potentially exploit this vulnerability, leading to script injection.