Podcast Generate

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

• Third-party content exposure detected (high risk: 1.00). The skill's generate.ts explicitly calls an external web search via zai.functions.invoke('web_search') and compiles searchResults (name, snippet, url) into the "material" string that is inserted into the LLM user prompt and used to generate the podcast script and TTS, so untrusted public web content can directly influence generation and subsequent actions.