High [CVE-2026-62663] Information Disclosure via Path Traversal in Media Filters

Banks generates meaningful LLM prompts using a simple template language. In versions prior to 2.4.4, all four media filters (image, audio, video, document) in banks accept untrusted user input as file paths via Path(value) and pass them directly to open(file_path, "rb") without any path sanitization, canonicalization, or directory restriction.

An attacker who controls template variables passed to a banks Prompt can use path traversal (../) to read arbitrary files accessible to the Python process—including.env files, SSH keys, cloud credentials, source code, /etc/passwd, and /etc/shadow—with the content returned base64-encoded in the rendered prompt output, making exfiltration trivial.

This is particularly dangerous for applications that use banks to process user-provided template variables before sending prompts to an LLM. This issue has been fixed in version 2.4.4.

A flaw was found in Banks. An attacker who can control template variables processed by Banks can exploit a path traversal vulnerability in the media filters (image, audio, video, document).

The content of these files can then be exfiltrated, leading to unauthorized information disclosure. This information disclosure flaw in the `banks` library allows an attacker to read arbitrary files via path traversal when processing untrusted user input in media filters.