game-marketing-context

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

• Third-party content exposure detected (high risk: 0.90). The skill's SKILL.md explicitly requires collecting and embedding verbatim player language from public, user-generated sources — e.g., "Verbatim Player Language Collection" listing Telegram group messages, itch.io comments, Reddit threads, and app store reviews — and then instructs using those exact quotes in generated marketing context files, meaning the agent will fetch/ingest untrusted third‑party content and act on it.