The Agent Skills Directory

[SAFE]: The skill implements significant security measures to protect user data during desktop interactions. It includes recurring instructions to treat recovered text from screens or command outputs as non-authoritative data, mitigating against potential indirect prompt injection from untrusted application content.
[SAFE]: Data privacy is addressed through mandatory redaction requirements for sensitive UI elements and explicit prohibitions against including credentials or private screen content in logs, chat, or receipts.
[SAFE]: The included helper script scripts/verify_preferences.py is implemented with security best practices, such as symlink traversal prevention, strict file permissions (0600) for configuration files, and atomic file writes using temporary locks.
[SAFE]: External dependencies and providers (e.g., Cua Driver, Playwright, Orca) are referenced for legitimate verification purposes with links to official documentation and repositories from known entities.

tk-app-verify