ad-creative

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

• Third-party content exposure detected (high risk: 1.00). The skill's SKILL.md explicitly requires researching and ingesting real audience conversations from public, user-generated platforms (e.g., "Step 0: Audience & Cultural Research" and the table of primary research platforms and "Run 3-5 WebSearch queries" with example site:twitter.com, reddit.com, nairaland.com, Facebook Groups, TikTok), instructing the agent to capture exact quotes and use those findings to define angles and generate ad creative—i.e., the agent must read and act on untrusted third‑party content.