claude-api

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

• Third-party content exposure detected (high risk: 0.70). The skill includes runtime flows that ingest untrusted third‑party content — e.g., vision inputs can be provided via URL ("Input methods: Base64 encoded, URL (if accessible)") and the Tool Use / Vision examples reference searchTool/saveTool and executing external APIs (e.g., fetchWeatherAPI, searchTool) which imply fetching public web/search results that the agent will read and interpret.