web-search

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

• Third-party content exposure detected (high risk: 0.85). At runtime, the skill fetches public web search results from outsider-operated providers (e.g., DuckDuckGo/MWMBL/Brave/Tavily/Perplexity) and normalizes their returned fields like title, url, and snippet into combined.json, which is then typically piped into the agent’s LLM context (indirect prompt injection via untrusted snippet/cited_text).