landing-page-guide

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

• Third-party content exposure detected (high risk: 0.90). This skill explicitly requires ingesting untrusted, user-generated content—e.g., "Pre-step: Research top 3-5 pain points from Reddit/communities" and "Design references: [include HTML code from inspiring sites]" in the SKILL.md workflow—so the agent is expected to read/interpret public web/forum content that can materially influence generated prompts, copy, and actions.