tmux

This tmux skill is largely benign in content: it documents controlling tmux sockets, sending keystrokes, and scraping pane output. There is no embedded network exfiltration, obfuscated code, hardcoded credentials, or download-and-execute patterns in the provided text. However, the capability it exposes—programmatic injection of keystrokes into interactive shells—is powerful and can be abused to execute arbitrary commands, read local secrets, or run third-party CLIs that perform further actions. The primary risks are operational: accidental or malicious execution of commands inside accessible tmux sessions, encouragement of autonomous agents to run uncontrolled actions (via '--yolo'), and unknown behavior of referenced external scripts. Recommendation: treat this skill as requiring high operational caution—ensure tmux socket directory permissions are restricted, only send trusted commands into sessions, audit any referenced helper scripts, and avoid running autonomous agents with unconstrained privileges inside tmux.