world-builder

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

• Third-party content exposure detected (high risk: 1.00). The SKILL.md explicitly requires using the socialdata MCP calls (e.g., get-profile, search-tweets) and web searches to fetch Twitter/X tweets, profiles, and public web content and directs the agent to quote and base NPC dialogue, map layout, and tool-call decisions on that user-generated public content, exposing the agent to untrusted third-party input.