The vulnerability in Style Dictionary arises from the convertTokenData utility function processing token arrays containing keys that manipulate the prototype chain, specifically keys like '{__proto__.foo}'. When such tokens are processed, they cause pollution of the global Object.prototype, effectively injecting properties globally. This can occur directly via convertTokenData(tokens, { output: 'object' }) or indirectly through the Expand API or the transform lifecycle, which synchronize token data back to objects. The vulnerability affects versions >=4.3.0 and <5.4.4. A patch fixing this issue was released in version 5.4.4. The vulnerability has been confirmed with a test and reproduction, and a pull request containing the fix is publicly available. Workarounds involve sanitizing token data to detect and reject keys containing '__proto__' and disabling the expand feature to prevent indirect exploitation.

Prototype pollution allows an attacker to modify the global Object prototype, potentially leading to arbitrary code execution, data corruption, or denial of service. The impact is high for NodeJS server applications using Style Dictionary, where malicious tokens can affect server behavior. For web applications, the impact is moderate, and for typical use cases where token data is controlled and access is limited, the impact is low.

A patch is available in Style Dictionary version 5.4.4. Users should upgrade to this version or later to remediate the vulnerability. As a workaround, sanitize token data before processing by recursively checking for keys containing '__proto__' and rejecting such tokens. Additionally, set the expand configuration option to false to prevent indirect exploitation via the Expand API. Example code for detection is provided in the advisory.