An authenticated user who can configure custom AI service URLs can cause the SparkyFitness server to perform outbound POST requests to arbitrary destinations without validation. The vulnerability allows the attacker to send requests with a controlled bearer token, probe internal services, and leak the AI system prompt, potentially exposing sensitive information and compromising internal network resources. This is a classic Server‑Side Request Forgery flaw that exploits the lack of destination validation in the AI flow code.

The vulnerability exists in CodeWithCJ SparkyFitness on all releases prior to version 1.6.0. Users running any pre‑1.6.0 build are affected; versions 1.6.0 and later include the fix that validates custom AI service URLs.

The CVSS score of 5.4 indicates a moderate severity assessment. With no EPSS data and no listing in the CISA KEV catalog, the likelihood and known exploitation activity are uncertain, but the impact range remains significant for authenticated users of the application. An attacker can compromise confidentiality of internal services and the AI prompt by using the unvalidated URL field, so the risk is moderate to high for exposed, public installations that allow custom AI services.