computer-use-agents

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

• Third-party content exposure detected (high risk: 0.75). The required runtime workflow captures the current desktop via pyautogui.screenshot() / scrot and feeds the resulting image (base64) into the agent’s LLM context; the screenshot is effectively free-form content from whatever is displayed (potentially outsider-authored web pages/messages/apps), creating indirect prompt-injection risk.

MEDIUM W021: Hidden or invisible Unicode characters detected (potential obfuscation or prompt injection).

• Hidden Unicode characters detected (1 type(s) found)