computer-use

CRITICAL E006: Malicious code pattern detected in skill scripts.

• Malicious code pattern detected (high risk: 0.90). The skill exposes stealthy background GUI automation (click/type/capture) plus screenshot capture and delivery, which enable high-risk abuse patterns — covert data exfiltration, credential harvesting, or remote control — despite the included safety guidance.

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

• Third-party content exposure detected (high risk: 0.75). The required runtime workflow uses computer_use(action="capture", mode="som", app=...) to ingest an AX-tree and screenshot overlays from the user’s currently driven GUI (e.g., a browser page), which can include outsider-authored free text from public web content or other third-party pages into the LLM context.