Desktop Control

Guidance

This skill appears to be what it claims: powerful desktop automation using pyautogui. Before installing or running it, consider the following: - Understand the power: it can move your mouse, type, press hotkeys, capture screenshots, and read/modify the clipboard — all of which can expose sensitive data or cause actions on your machine. This is expected behavior for a desktop-automation skill, not a hidden backdoor. - Prefer running in a safe environment: test in a disposable VM, non-production account, or on a system without sensitive documents open. Close important apps before running demos. - Use safety options: enable failsafe (move mouse to corner to abort) and set require_approval=True if you want manual confirmation for each action. Review demos before running them. - Beware of autonomous invocation: if you allow the agent to invoke skills autonomously, it could run sequences without your interactive confirmation. If you do not trust the skill/user code, disable autonomous invocation or only allow manual/user-invoked runs. - Review the code: if you plan to run this long-term, inspect the full ai_agent.py and any truncated parts for network calls or code that might upload screenshots/clipboard data. The provided fragments show no network exfiltration, but the files were truncated in places — verify the remaining code sections yourself. - Install dependencies from official sources and avoid running unknown binaries. If you need to grant broader privileges (e.g., run as admin), reconsider usage. If you want, I can scan the remaining truncated portions (full ai_agent.py and __init__.py) for any network calls, hidden endpoints, or suspicious behaviors to raise confidence further.