cmux Control (claw-mux)

What it does

This skill provides programmatic control over cmux terminal topology and terminal I/O. It lets an agent identify the current workspace, split panes, send commands to specific surfaces, read screen output (including scrollback), and manage notifications/status in the cmux sidebar. The skill documents completion detection strategies (wait-for, content polling) and supplies helper scripts for polling and synchronization.

When to use it

Use when automating terminal workflows inside a cmux environment: running builds, starting servers and waiting for a 'ready' marker, orchestrating multi-pane test runs, or streaming logs from other panes. Falls back to standard tools when not running inside cmux.

What's included

• Scripts: poll-screen helpers and other automation scripts (has_scripts=true).
• References: terminal I/O patterns, sync-and-automation, handles and identify docs (has_references=true).
• Instructions: examples for send/read patterns, wait-for usage, and common gotchas (session-scoped short refs, env checks).

Compatible agents

Best with agents that can execute shell commands and read terminal output (Claude Code, local CLI agents, GitHub Actions runners when cmux is available).