Linux Omniscient

This skill indeed contains code to control many parts of a host (camera, microphone, screen, mouse/keyboard automation, network settings, power actions, process management). That aligns with its stated purpose but has three practical concerns: (1) multiple coding inconsistencies (e.g., dependency list includes Windows-only packages while scripts are Linux-focused; many run_cmd calls pass lists or include pipe characters as list elements — these will often fail), (2) it can perform high-risk actions (shutdown, suspend, change DNS/proxy, access camera/screenshots, kill processes, send HTTP requests with tokens), and (3) SKILL.md advocates automatic orchestration for 'any task', which could let the agent chain many sensitive operations without clear, enforced user confirmations. Before installing or enabling this skill: - Run it only in a sandbox or VM first (do not enable on a production or sensitive machine). - Review and fix the code (the common.run_cmd contract doesn't match many callers; shell/pipe handling is incorrect) or ask the author to provide a tested, platform-consistent release. - Require explicit, per-operation user confirmation for high-risk actions (shutdown/restart, network changes, camera/microphone access, sending tokens to remote endpoints). - If possible, disable autonomous invocation or restrict the skill to manual invocation until you trust it. - If you must use IoT/HomeAssistant features, avoid passing long-lived tokens to the skill unless you trust its runtime policies; prefer short-lived credentials. - Ask the publisher to clarify Windows vs Linux target platforms and to clean up dependency and command handling. Given the broad host control surface and the implementation sloppiness, treat this skill as risky until cleaned up and run under tight controls.