How exploitable is this vulnerability in practice?

It is highly exploitable in environments where users are allowed to submit or edit evaluator code. Any authenticated user with such access can typically craft payloads that leverage `numpy.ma.core.inspect` to escape the sandbox and run code on the server.

What level of access does an attacker need?

The attacker must have authenticated access to the Agenta self-hosted API with permissions to define or execute custom evaluator code. No direct OS-level access is required; exploitation occurs entirely via the application layer.

What is the safest general fix pattern?

The most robust pattern is to avoid running untrusted user code in the same interpreter process as the main application. Instead, run user code in isolated execution environments (e.g., containers or VMs) with strict OS-level sandboxing and minimal privileges, exposing only a controlled API surface.

Severity:

HIGH RISK

(8.8)

Improper Code Execution Sandbox (CWE-94) in Agenta Custom Code Evaluator

CVE-2026-27952

Published:

Feb 25, 2026

Status:

Analyzed

Summary

Affected Context

Exploit Preconditions

Why This Happens

Potential Impact

General Mitigation Guidance

How Teams Detect Issues Like This

Frequently Asked Questions

Source

NDV

Recent Vulnerabilities

CWE-78

CRITICAL RISK

CRITICAL

(9.9)

Command Injection in MCP Stdio Configuration Validation (CWE-78) in WeKnora

Remote Code Execution

March 7, 2026

CWE-89

CRITICAL RISK

CRITICAL

(9.9)

SQL Injection via PostgreSQL Array/Row Expressions (CWE-89) in WeKnora Query Engine

Remote Code Execution via SQL Injection on the Database Server

March 7, 2026

Vulnerability Type

Category:

Code Injection / Sandbox Escape

CWE:

CWE-94

Impact:

Remote Code Execution within the Agenta API server process

Severity:

HIGH RISK

(8.8)

CVSS Version:

3.1

Vulnerability Type

Category:

Code Injection / Sandbox Escape

CWE:

CWE-94

Impact:

Remote Code Execution within the Agenta API server process

Severity:

HIGH RISK

(8.8)

CVSS Version:

3.1

Catch This CVE While Code Is Being Written

Detect vulnerable code directly in pull requests, so CVEs are fixed before merge.

Get Started

Ship clean & secure code faster

START 14 DAYS FREE TRIAL

BOOK A DEMO

Made with Love in San Francisco

355 Bryant St. San Francisco, CA 94107, USA

Ask AI for summary of CodeAnt

Made with Love in San Francisco

355 Bryant St. San Francisco, CA 94107, USA

Ask AI for summary of CodeAnt

Made with Love in San Francisco

355 Bryant St. San Francisco, CA 94107, USA

Ask AI for summary of CodeAnt

Improper Code Execution Sandbox (CWE-94) in Agenta Custom Code Evaluator

Summary

Affected Context

Exploit Preconditions

Why This Happens

Potential Impact

General Mitigation Guidance

How Teams Detect Issues Like This

Frequently Asked Questions

Source

Recent Vulnerabilities

CRITICAL RISK

CRITICAL

(9.9)

Command Injection in MCP Stdio Configuration Validation (CWE-78) in WeKnora

CRITICAL RISK

CRITICAL

(9.9)

SQL Injection via PostgreSQL Array/Row Expressions (CWE-89) in WeKnora Query Engine

Vulnerability Type

Vulnerability Type

Vulnerability Type

Catch This CVE While Code Is Being Written

Ship clean & secure code faster

Product

Pricing

Company

Legal

Developers

Compare

Product

Pricing

Company

Legal

Developers

Compare

Product

Pricing

Company

Legal

Developers

Compare