CVE Index中文
L3 · ShartedNode.js · npm · Code injectionpromzard-context-code-injection

PromZard can execute JavaScript supplied through context property names

CERT/CC disclosed CVE-2026-127109 in promzard 4.0.0 and earlier. When a caller accepts attacker-controlled property names in the template context object, loading even a trusted template can execute JavaScript with the privileges of the Node.js process.

Disclosed
Nov 12, 2026
Public disclosure date
Severity
Sharted
1 CVE tracked here
Splash zone
Applications and tools using promzard 4.0.0 or earlier with context property names an attacker can influence.
Node.js · npm · Code injection
In the wild
Unknown
The CERT/CC disclosure does not report confirmed exploitation.

What happened

PromZard evaluates prompt templates with a caller-supplied context object. Its template loader uses the property names as formal parameters in generated JavaScript, then evaluates that source with Node.js. An attacker who controls a property name can inject code even when the template is trusted. Trace how each caller constructs the context to determine whether an attacker can reach that path.

This is L3 because it affects a widely distributed npm package and can lead to code execution in vulnerable callers. The npm registry still listed 4.0.0 as the latest published release on November 14, 2026. Check the registry and upstream guidance before relying on any subsequent upgrade.

What’s in the bowl

JavaScript code injection through context property names

The template loader interpolates context property names into JavaScript source and evaluates it in the current Node.js context.

Crime-scene timeline

  1. Oct 13, 2026

    CERT/CC notifies PromZard

    CERT/CC lists this as the vendor notification date and says it did not receive a statement.

  2. Nov 12, 2026

    CERT/CC publishes VU#685399

    The disclosure identifies affected releases up to and including 4.0.0 and advises callers to validate context keys.

Who said what

We list each claim and who made it. Primary sources (the vendor, the researcher, the registry) are marked. Where reports disagree, we grade by the worst credible case.