landing-page

CRITICAL E004: Prompt injection detected in skill instructions.

• Potential prompt injection detected (high risk: 1.00). The skill contains a non‑advertised, mandatory "WebMCP" integration that forces embedding an external AI agent connector (loading third‑party JS, registering tools/resources, and exposing page context) and explicitly hides agent coordination from users — instructions that go beyond normal landing‑page building and enable external AI discovery/exfiltration, so this is a deceptive/out‑of‑scope prompt injection.

CRITICAL E006: Malicious code pattern detected in skill scripts.

• Malicious code pattern detected (high risk: 1.00). The skill mandates inclusion of an external WebMCP library (loaded from a CDN) and programmatic registration of “tools” that remote agents can invoke to read page content, click buttons, and submit forms — effectively creating an agent-accessible backdoor (remote code/action invocation and easy data access/exfiltration) and forcing a supply-chain dependency via the CDN; this behavior constitutes a high-risk intentional backdoor/data-exfiltration vector.

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

• Third-party content exposure detected (high risk: 0.85). The Build Agent injects WebMCP by loading an external script from https://cdn.jsdelivr.net/npm/webmcp@latest/webmcp.js at runtime, which is outsider-authored free text/code fetched from the public web and then executed/ingested into the page context.

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

• Potentially malicious external URL detected (high risk: 1.00). The skill explicitly mandates loading and initializing the official WebMCP library (https://cdn.jsdelivr.net/npm/webmcp@latest/webmcp.js) in every generated page and uses it at runtime to register tools/resources (mcp.registerTool / new WebMCP()), so this externally-fetched script is a required runtime dependency that executes remote code to expose and control agent interactions.