The Agent Skills Directory

[COMMAND_EXECUTION]: The skill leverages tmux send-keys to perform automated command execution within terminal panes. This allows an agent to drive shell sessions and execute arbitrary code.
Evidence: The command tmux send-keys -t work:1.0 "ls -la" Enter and the scripting layout example demonstrate how to send and execute commands programmatically.
[INDIRECT_PROMPT_INJECTION]: The skill utilizes tmux capture-pane to ingest terminal content into the agent's context. This ingestion point is vulnerable to indirect prompt injection if the terminal has processed attacker-controlled data (e.g., file contents or server logs).
Ingestion point: tmux capture-pane -t work:1.0 -p (SKILL.md).
Capability inventory: The skill has command execution capabilities (send-keys) and file system write access (save-buffer, pipe-pane).
Boundary markers: None present; there are no instructions to use delimiters or ignore embedded commands when processing captured output.
Sanitization: No sanitization or validation of the captured output is described before it is utilized by the agent.

tmux