perplexity

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

• Third-party content exposure detected (high risk: 0.90). The skill explicitly performs web searches and fetches remote page/file/image contents (see references/pro-search.md for built-in web_search and fetch_url_content and references/media.md/references/search-api.md for file_url/image_url and search results), so the agent ingests untrusted public web/user-generated content that can influence tool orchestration and subsequent actions.

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

• Potentially malicious external URL detected (high risk: 0.90). Pro Search includes a runtime tool fetch_url_content that fetches arbitrary external pages (example: "https://arxiv.org/pdf/...") and injects their contents into the model's reasoning/response stream, meaning fetched URL content can directly control the agent's prompts/context at runtime.