A path traversal flaw in the Beifong AI Podcast Agent lets an unauthenticated attacker request a crafted path that bypasses the intended filesystem boundaries and reads any file the server process can access. The vulnerability originates in the stream‑audio endpoint, which concatenates a user‑controlled parameter directly into a file path without validation. This can lead to disclosure of sensitive data such as configuration files, credentials, or other private information stored on the server.

The vulnerability affects the awesome_llm_apps project maintained by TheUnwindAI. Any release containing the vulnerable commit (e46690f99c3f08be80a9877fab52acacfab8251, dated 2026‑01‑19) is potentially impacted. No additional version restrictions are listed.

The CVSS score of 7.5 indicates a high severity, but the EPSS value of less than 1% suggests that exploitation is currently unlikely. The flaw is not listed in CISA’s KEV database, implying no known public exploits at this time. The likely attack vector is an unauthenticated HTTP request to the stream‑audio endpoint with a crafted path argument that allows an attacker to read arbitrary files on the server.