What problem does it solve? When an AI agent runs inside a cmux terminal, it can accidentally act on the wrong workspace, steal the user's focus, or disrupt the layout the user is looking at. This Skill anchors every cmux command to the workspace and surface that invoked the agent, so automation stays non-disruptive and auditable. ## Core Features & Use Cases - Caller-scoped automation: Resolves the current workspace and surface via CMUX_WORKSPACE_ID, CMUX_SURFACE_ID, and cmux identify, then targets all mutations with explicit --workspace and --surface flags. - Non-disruptive layout building: Creates panes and surfaces additively with new-pane and new-surface, passing --focus false and avoiding focus-changing verbs like select-workspace or focus-pane. - Sidebar status and logging: Attaches status, progress bars, and log entries to the current workspace so the sidebar reflects the running task. - Use Case: While running a dev server, ask the agent to open a browser preview in a right-side helper pane and post build progress to the sidebar, all without moving your cursor or switching your active tab. ## Quick Start Ask the agent to identify the current cmux workspace and open a browser preview of localhost:3000 in a right-side pane without changing focus.
