LiteLLM, a proxy server for calling LLM APIs, prior to version 1.83.7-stable, did not properly validate file paths during extraction of uploaded skill ZIP archives. Authenticated users with access to LiteLLM LLM API routes or keys permitting /v1/skills, anthropic_routes, or llm_api_routes could exploit this to perform path traversal, causing files to be written outside the designated extraction or staging directories. This vulnerability is identified as CWE-22 and has a CVSS 4.0 score of 6.1 (medium severity). The issue is resolved in version 1.83.7-stable.

An authenticated user with specific API route permissions can upload a malicious skill archive that writes files outside the intended directories, potentially overwriting or creating files in unauthorized locations. This could lead to unauthorized file modification or code execution depending on the environment and usage context. No known exploits are reported in the wild.

Upgrade to LiteLLM version 1.83.7-stable or later, where the path traversal vulnerability is fixed. Since no official remediation level or patch link is provided, rely on the vendor's version update as the fix. Until upgraded, restrict access to API routes /v1/skills, anthropic_routes, and llm_api_routes to trusted users only.