CVE-2026-55791

1. Overview

Craft CMS is vulnerable to Server-Side Request Forgery (SSRF) and Arbitrary JavaScript Injection through the /actions/app/resource-js endpoint. By exploiting the default permissive trustedHosts configuration, an attacker can poison the Host or X-Forwarded-Host header to manipulate the application’s $baseUrl. This bypasses the endpoint’s internal URL validation, forcing the backend Guzzle client to fetch a malicious payload from an attacker-controlled server and reflect it to the client with a Content-Type: application/javascript header.

2. Vulnerability Mechanism (Root Cause)

The vulnerability manifests when assetManager.cacheSourcePaths is set to false. The attack chain relies on three structural flaws and insecure defaults:

• A. Default Proxy Trust (trustedHosts): Craft’s default GeneralConfig::$trustedHosts is set to ['any']. This allows an attacker to bypass front-end web server (Nginx/Apache) strict Host header validations by simply injecting an X-Forwarded-Host header. Yii2 will parse this and globally set $baseUrl to the attacker's domain.
• B. Insecure HTTP Client (actionResourceJs): In AppController::actionResourceJs(), the strstartswith($url, $baseUrl) validation is bypassed because $baseUrl is already poisoned by the attacker. The core then uses Craft::createGuzzleClient()->get($url). Unlike the GraphQL Asset fetcher, this Guzzle instance defaults to ALLOW_REDIRECTS => true.
• C. Forced JS Content-Type: The response fetched from the attacker's server is blindly returned to the user via $this->asRaw() with the header Content-Type: application/javascript.

3. Attack Scenario & Impact (Proof of Exploitability)

This endpoint acts as a proxy, taking remote, unverified content and serving it as valid JavaScript. While the direct SSRF allows for internal network probing, the most devastating impact occurs when caching layers are involved.

If the Craft CMS instance is behind a caching layer, this vulnerability leads directly to Web Cache Poisoning:

1. An unauthenticated attacker sends the poisoned request.
2. The caching layer caches the malicious JavaScript response for the legitimate /actions/app/resource-js URI.
3. When an authenticated Administrator logs into the Control Panel, their browser loads the poisoned cached JavaScript (Stored XSS).
4. The malicious script extracts window.Craft.csrfTokenValue and silently sends a POST request to /admin/actions/plugins/install-plugin, achieving 1-Click Remote Code Execution (RCE) via Session Riding.