grok

CRITICAL E005: Suspicious download URL detected in skill instructions.

• Suspicious download URL detected (high risk: 0.80). These URLs include a direct shell installer (curl -fsSL https://x.ai/cli/install.sh | bash) and an example S3 upload endpoint (https://s3.example.com) — piping remote .sh to a shell and relying on S3-hosted presigned URLs are common vectors for malware distribution unless you explicitly trust and verify the publisher and script contents.

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

• Third-party content exposure detected (medium risk: 0.30). In the required runtime workflow, outsider-authored free text is ingested when Grok runs X “x_search” commands (including full post text via thread/x_thread_fetch) and live web research via web_fetch/web_search, because Grok server-side reads user-selected posts/pages and returns them into the model’s answer.