web-search

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

• Third-party content exposure detected (high risk: 0.85). The skill performs runtime web search via external providers (e.g., DuckDuckGo/Bing/Google/Searxng) and ingests their returned JSON fields like snippet/content into the agent’s tool output, which is then available to the agent as LLM context; these snippets are outsider-authored web content.