The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes tmux commands to manage terminal sessions, capture output, and send keystrokes. These are the core intended functions of the skill.
[DATA_EXPOSURE]: The skill accesses terminal output via tmux capture-pane. This is necessary for its stated purpose of scraping pane output and does not involve unauthorized file access or network transmission.
[REMOTE_CODE_EXECUTION]: While the documentation mentions running coding agents like codex or claude-code, the skill itself does not download or execute remote code; it provides a framework for the user to orchestrate their own local tools.
[INDIRECT_PROMPT_INJECTION]: The wait-for-text.sh script processes terminal output which could technically contain adversarial text. However, the script only performs regex matching for control flow and does not feed this data back into an LLM prompt in an unsafe manner within the provided scripts.

tmux