The h3 HTTP framework uses a simple startsWith check in its mount() method to decide whether a request falls under a mounted sub‑application’s path prefix. This check does not verify that the next character after the base is a slash or the end of the string, so middleware registered on a path such as "/admin" will also run for unrelated routes such as "/admin-public", "/administrator", or "/adminstuff". The result is that an attacker can trigger context‑setting or other middleware on paths it was never intended to cover, potentially polluting the request context with unintended privilege flags.

Node.js applications that import the h3 package and use the mount() feature are affected when the package version is between 2.0.0-0 and 2.0.1-rc.16 inclusive. All release candidates from rc1 through rc16 of the 2.0.1 series are listed as vulnerable. The vulnerability is fixed in version 2.0.2-rc.17 and newer.

The CVSS score of 3.7 indicates moderate severity, while an EPSS less than 1% suggests a low likelihood of exploitation. The vulnerability is not listed in CISA’s KEV catalog. The likely attack vector involves an attacker sending HTTP requests to the target application and knowing or guessing a mounted path prefix. Based on the description, it is inferred that an attacker could alter request context fields, which may affect authorization decisions. The impact depends on how the application uses the context. Overall risk is moderate.