Spring AI LLM-Influenced Filename Vulnerability in Anthropic Skills API Support

A vulnerability exists in Spring AI versions 1.1.0 through 1.1.x, specifically in the support for Anthropic's Skills API. The issue arises because LLM-influenced filenames are used unsanitized in the Path.resolve method before files are written to disk. This flaw could enable a malicious user to write files outside the designated target directory, potentially including restricted directories.

Impact

Exploitation of this vulnerability could lead to unauthorized file writing in restricted directories, bypassing intended directory limitations.

Remediation

Users of affected Spring AI versions should upgrade to version 1.1.7.