The CVE describes an improper certificate validation flaw in the WebSocket streaming client of the agentjido req_llm library. Because the client ignores TLS certificate verification, an attacker on the network path can present a forged certificate and intercept the WebSocket connection. The bearer token or OAuth access token is transmitted in the Authorization header, so the attacker can read and modify it, effectively compromising authentication credentials and enabling unauthorized access to OpenAI services. This weakness is a classic Improper Certificate Validation (CWE‑295) and undermines both confidentiality and integrity of the credentials.

Applications that depend on agentjido req_llm versions 1.9.0 through 1.28.x are vulnerable. The library’s WebSocket paths are used for OpenAI Responses, Realtime, and Codex streaming, so any deployment that invokes these endpoints within the affected version range is impacted. Systems that embed the library for streaming or realtime OpenAI interactions are thus at risk.

The CVSS score of 9.1 categorizes the vulnerability as critical. Although an EPSS score is not available, the exploitability remains high because the flaw requires only a TLS man‑in‑the‑middle capability, a common scenario for on‑path adversaries. The vulnerability is not currently listed in the CISA KEV catalog, but the lack of certificate verification directly exposes bearer tokens. Based on the description, the attack vector is inferred to be a network‑based MITM that intercepts the insecure WebSocket traffic.