agnes-free-text

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (high risk: 0.70). Outsider free text can be injected via --message role:content / --prompt / --system arguments (user-supplied text not authored by the operating user is passed verbatim into payload["messages"] and sent to the Agnes API, which then feeds it into the model context).