vp-interaction-routing

CRITICAL E006: Malicious code pattern detected in skill scripts.

• Malicious code pattern detected (high risk: 0.90). This bridge intentionally exposes native UI automation and app state (including optional screenshots) to external MCP clients without built-in intent enforcement (auto-approval can be enabled), creating a high-risk capability that can be used for data exfiltration and remote control of the host.

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

• Third-party content exposure detected (medium risk: 0.30). The required workflow in SKILL.md routes “web-page content or interaction” to browser tooling while treating “page content as untrusted data,” meaning outsider-authored text can enter the agent’s runtime via whatever browser page the user requests it to interact with.