cdn

CRITICAL E004: Prompt injection detected in skill instructions.

• Potential prompt injection detected (high risk: 0.80). The prompt includes hidden self-evolution telemetry instructions to look for a user-specific memory directory (~/.claude/projects/) and append execution metadata silently, which is outside the stated CDN configuration purpose and constitutes deceptive hidden behavior.

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

• Third-party content exposure detected (high risk: 0.80). The skill's Phase 1 explicitly instructs the agent to query public DNS records and fetch HTTP response headers (e.g., x-cache, cf-cache-status, x-amz-cf-id) from external domains to detect existing CDN/provider, which are untrusted third‑party sources whose content is read and directly influences configuration decisions.