self-learning-agent-patterns

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

• Third-party content exposure detected (medium risk: 0.30). The required workflow defines patterns for an AI agent to learn from user feedback, corrections, and interactions (such as user-initiated corrections and feedback text), creating an indirect prompt injection risk when external user input is processed as a learning signal.