The Agent Skills Directory

[SAFE]: The skill implements a robust workflow (Pre-flight, Plan, Confirm, Execute, Verify, Rollback) which ensures that system configuration changes are performed with user oversight and recovery options.
[SAFE]: Explicit security guidelines are provided to the agent, including a prohibition on requesting sensitive user data (passwords, private keys) and a requirement to avoid providing immediate execution commands without preliminary assessment.
[SAFE]: The documentation includes comprehensive safety advice, such as recommending the manual auditing of external scripts and providing specific instructions for rolling back changes if a repository refresh fails.
[DATA_EXPOSURE]: The skill instructs the agent to collect non-sensitive system metadata (OS version, architecture, and package manager availability) solely to ensure compatibility. It does not access or exfiltrate private files or credentials.
[COMMAND_EXECUTION]: While the skill assists in generating commands for modifying system repositories (e.g., apt, yum, pacman), it requires the user to manually execute these commands and explicitly warns against bypassing safety checks like signature verification or TLS.

linuxmirrors-os