web-search

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

• Third-party content exposure detected (high risk: 0.95). The required runtime workflow uses ddgr to fetch DuckDuckGo search results (public web content) and then uses WebFetch to retrieve the selected result page body, which is outsider-authored free text injected into the agent’s LLM context.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.90). This skill runs ddgr to obtain search-result URLs and then uses WebFetch at runtime to retrieve and inject external page content (e.g., https://nextjs.org/docs/app/building-your-application/authentication) into the agent's context to synthesize replies, so fetched URLs can directly control prompts.