perplexity

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

• Third-party content exposure detected (high risk: 1.00). The skill explicitly performs web searches and retrieves arbitrary public URLs (e.g., Pro Search's built-in web_search and fetch_url_content in references/pro-search.md, search.create in SKILL.md/references/search-api.md, and file_url/image_url handlers in references/media.md) so the agent ingests untrusted, user-provided third-party web content that can influence subsequent reasoning and tool use.

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

• Potentially malicious external URL detected (high risk: 0.90). Pro Search's built-in fetch_url_content tool fetches and injects remote page content at runtime (example URL: https://arxiv.org/pdf/...), which directly influences the model's reasoning/output and is required for Pro Search workflows, so it meets the criteria for a runtime external dependency that can control prompts.