The Agent Skills Directory

[SAFE]: The skill implements a robust safety framework for terminal commands, explicitly requiring human confirmation for irreversible or externally visible actions such as data deletion, force-pushes, database modifications, and deployments.
[SAFE]: Includes specific instructions to mask sensitive token-like values in command output before presenting them to the user.
[SAFE]: Safely handles unattended execution environments (cron jobs, one-shot runs, webhooks) by defaulting to a 'deny' posture for flagged commands and directing them to a human review turn.
[SAFE]: Establishes a clear 'prepared-vs-observed' boundary, preventing the agent from claiming success or execution without verifiable evidence from the terminal runtime.

omh-terminal