Command Injection

Overview PraisonAI is a PraisonAI is an AI Agents Framework with Self Reflection. PraisonAI application combines PraisonAI Agents, AutoGen, and CrewAI into a low-code solution for building and managing multi-agent LLM systems, focusing on simplicity, customisation, and efficient human-agent...

CVE-2026-61613

Cursor: Cloud Agent browser sandbox escape where attacker-controlled web content inside the agent container could reach an unauthenticated local agent endpoint, allowing code execution and access to files, repository contents, environment variables, credentials, and GitHub App tokens. Fixed 2026-...

CVE-2026-61613 Cursor: Cloud Agent Browser Sandbox Escape

Cursor is a code editor built for programming with AI. Prior to the Cloud Agent fix on 03/31/2026, browser-enabled Cursor Cloud Agent sessions allowed attacker-controlled web content to connect from inside the agent container to an unauthenticated local agent endpoint, enabling code execution...

CVE-2026-46635

A flaw was found in Twig, a template language for PHP. An untrusted template author, with the column filter in their allowed filters, can exploit this vulnerability. The column filter incorrectly passes object arrays to PHP's arraycolumn function, which can read properties that are not restricted...

CVE-2026-48808

A flaw was found in Twig, a template language for PHP. The column filter in Twig fails to properly forward the active sandbox state, specifically the current Source to the SandboxExtension::checkPropertyAllowed function. This oversight results in the loss of SourcePolicyInterface decisions,...

CVE-2026-48805

A flaw was found in Twig, a template language for PHP. Deprecated internal wrappers in the src/Resources/core.php file do not correctly forward the sandbox state. This allows legacy function calls, such as twigarraysome, to bypass security restrictions imposed by the sandbox. An attacker could...

CVE-2026-49981

A flaw was found in Twig, a template language for PHP. An attacker with low privileges could exploit a caching issue in the template sandboxing mechanism. This allows a sandboxed template to bypass its intended security policy, potentially leading to unauthorized information disclosure and limite...

CVE-2026-46639

A flaw was found in Twig, a template language for PHP. An attacker with write access to a sandboxed Twig template can bypass security restrictions due to an issue in object-destructuring assignment. This allows them to read public properties or invoke public methods on objects, leading to...

CVE-2026-46638

A flaw was found in Twig, a template language for PHP. A security vulnerability exists where a template included within a restricted sandbox environment can bypass security checks if it was previously loaded without these restrictions. This allows the template to execute unauthorized functions,...

CVE-2026-48807

A flaw was found in Twig, a template language for PHP. The security sandbox, designed to restrict operations, does not properly handle certain values when processing templates. This oversight could allow an attacker to bypass security policies, potentially leading to the disclosure of sensitive...

CVE-2026-47732

A flaw was found in Twig, a template language for PHP. This vulnerability allows a sandboxed template author to bypass security policies by triggering PHP string coercion on specific language constructs. This can lead to information disclosure, as an attacker may be able to invoke toString on...

CVE-2026-61438

PraisonAI before 4.6.78 contains a remote code execution vulnerability in JobWorkflowExecutor.execinlinepython due to insufficient AST validation of workflow script steps. Attackers can create malicious YAML workflow files with import os statements followed by os.system calls that bypass sandbox...

CVE-2026-60085

PraisonAI before 4.6.78 contains an unenforced security policy vulnerability in the default Subprocess Sandbox backend where blockedcommands, blockedpaths, blockedimports, allowsubprocess, and allowfilewrite restrictions are completely ignored. Attackers can execute arbitrary subprocess commands,...

CVE-2026-61438

PraisonAI before 4.6.78 contains a remote code execution vulnerability in JobWorkflowExecutor.execinlinepython due to insufficient AST validation of workflow script steps. Attackers can create malicious YAML workflow files with import os statements followed by os.system calls that bypass sandbox...

CVE-2026-61438 PraisonAI before 4.6.78 Remote Code Execution via Broken AST Sandbox

PraisonAI before 4.6.78 contains a remote code execution vulnerability in JobWorkflowExecutor.execinlinepython due to insufficient AST validation of workflow script steps. Attackers can create malicious YAML workflow files with import os statements followed by os.system calls that bypass sandbox...

CVE-2026-61438 PraisonAI before 4.6.78 Remote Code Execution via Broken AST Sandbox

PraisonAI before 4.6.78 contains a remote code execution vulnerability in JobWorkflowExecutor.execinlinepython due to insufficient AST validation of workflow script steps. Attackers can create malicious YAML workflow files with import os statements followed by os.system calls that bypass sandbox...

EUVD-2026-44643

PraisonAI before 4.6.78 contains a remote code execution vulnerability in JobWorkflowExecutor.execinlinepython due to insufficient AST validation of workflow script steps. Attackers can create malicious YAML workflow files with import os statements followed by os.system calls that bypass sandbox...

EUVD-2026-44636

PraisonAI before 4.6.78 contains an unenforced security policy vulnerability in the default Subprocess Sandbox backend where blockedcommands, blockedpaths, blockedimports, allowsubprocess, and allowfilewrite restrictions are completely ignored. Attackers can execute arbitrary subprocess commands,...

CVE-2026-60085 PraisonAI before 4.6.78 Unenforced Security Policy in Subprocess Sandbox

PraisonAI before 4.6.78 contains an unenforced security policy vulnerability in the default Subprocess Sandbox backend where blockedcommands, blockedpaths, blockedimports, allowsubprocess, and allowfilewrite restrictions are completely ignored. Attackers can execute arbitrary subprocess commands,...

CVE-2026-60085 PraisonAI before 4.6.78 Unenforced Security Policy in Subprocess Sandbox

PraisonAI before 4.6.78 contains an unenforced security policy vulnerability in the default Subprocess Sandbox backend where blockedcommands, blockedpaths, blockedimports, allowsubprocess, and allowfilewrite restrictions are completely ignored. Attackers can execute arbitrary subprocess commands,...