agent-development

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

• Third-party content exposure detected (medium risk: 0.30). The skill documentation describes an agent system where the LLM consumes the user-provided “description” examples and system prompt loaded from the agent markdown (which can include arbitrary free text from the provider user at runtime via agent triggering and tool-enabled reading), so outsider-authored text can influence the agent without pre-selecting a specific item.