refine-dev

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

• Third-party content exposure detected (high risk: 0.80). The skill explicitly fetches and ingests arbitrary external API and realtime/websocket data (see references/inferencer.md "Inferencer fetches data from your API" and references/data-providers.md customDataProvider using fetch(API_URL) and realtime.md WebSocket/Ably examples), and that fetched, potentially untrusted content is parsed and used to generate UI/code and drive behavior (inferencer, hooks, live updates), so third-party content can materially influence the agent's actions.